#FPF Pattern-Quality Evaluation CharacteristicSpace

E.21 · Stable · Definitional pattern · Normative · Part E - The FPF Constitution and Authoring Guides

#About this pattern

This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.

#How to use this pattern

Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.

Type: Definitional pattern Status: Stable Normativity: Normative

Use E.21 when you are evaluating one FPF pattern version and need to know whether improvement can stop without relying on taste, a single quality score, or template compliance alone. The governed object is the scoped pattern-quality claim; the pattern version under quality read is one authored FPF pattern body in one declared use and reader scope.

#Relations

#Content

#Problem frame

Use E.21 when you are evaluating one FPF pattern version and need to know whether improvement can stop without relying on taste, a single quality score, or template compliance alone. The governed object is the scoped pattern-quality claim; the pattern version under quality read is one authored FPF pattern body in one declared use and reader scope.

Use it especially when a pattern already has the E.8 skeleton and may even pass an E.19 review, but authors still disagree about whether the pattern is recognisable, action-guiding, ontologically precise, SoTA-bearing, proportionate in apparatus, and safe to compose with neighbours.

Not this pattern when. Use E.8 to write the pattern body. Use E.19 to run admission or refresh review profiles. Use E.9.DA when the evaluated object is one DRR decision-adequacy claim rather than one authored FPF pattern version. Use C.16, A.17, A.18, and A.19 when the live problem is measurement legality, Characteristic and Scale discipline, or declaring a general CharacteristicSpace. Use C.25 when the live problem is an arbitrary engineering quality family rather than FPF pattern quality. Use F.18 when the live problem is durable naming. Not this pattern when the live question is whether a user correctly applied an FPF pattern in a project case. E.21 reads the quality of the pattern text for a declared reader, use, and scope. Pattern-application quality belongs to the receiving project-side or publication-side pattern that governs the actual case: evidence, assurance, gate, work, decision, publication, action invitation, method, or bridge use as applicable.

First useful move. The architectural first move of E.21 is not "run quality control". It is to recover the pattern version's first admissible action-guiding move for the declared reader, use, and window. Start with the ordinary first-pass slice of PatternQualityQBundle: name PatternVersionRef, WorkingReaderScope, IntendedUse, and QualificationWindow, then inspect the pattern version's Problem frame and Solution for one first admissible action-guiding move for that working reader.

If that first move is absent, unrecoverable, or only present in the Conformance Checklist, close the first-pass read as repairBeforeUse for the declared use, or as admissibleWithNarrowerUse when the pattern can still serve an expert-only or support-only scope. Do not require a full coordinate comparison before this result.

Create the fuller PatternQualityQBundle only when the first-pass slice survives, when several candidate edits are being compared, or when admission, refresh, high-assurance reuse, or contested neighbour authority is live.

What goes wrong if missed. Pattern improvement collapses into a single "quality score", a reviewer preference, or a heading checklist. That hides hard blockers such as undefined vocabulary, shadow authority, decorative SoTA, ordinal arithmetic, missing first move, or neighbour breakage.

What this buys. E.21 gives authors one typed quality space for FPF patterns: hard eligibility filters first, multi-characteristic coordinates second, Pareto-front reasoning for candidate improvements, and an explicit stop condition that does not scalarize.

E.21 is the pattern-quality evaluation pattern for claims of the form "this pattern version is sufficiently good for this use, readership, and scope." It normalizes such claims into a scoped bundle; it does not run the review process or take over authoring, measurement, naming, project evidence, assurance, gate, release, or work authority from neighbouring FPF patterns.

Governed object in plain terms. The governed object is the pattern-quality claim for one FPF pattern version under one declared use and reader scope.

Primary working reader. The first reader is an FPF author, reviewer, or steward deciding whether to keep improving a pattern version. The downstream reader is the practitioner or manager who must use the pattern as action guidance.

#Problem

FPF needs a way to evaluate pattern quality without creating a fake scalar. A single score is attractive because it is easy to compare, but it is false for this object. A pattern can be excellent in SoTA support and still unreadable; precise in ontology and still too heavy for ordinary use; concise and still wrong about neighbours.

The recurring failures are:

1. Hidden scalarization. Different coordinates such as recognition, ontology, SoTA, evidence, and reader cost are averaged or ranked as if they had one common unit.
2. Template-only maturity. Canonical headings are present, but the first working situation, first admissible move, boundary, and practical payoff remain missing.
3. Checklist substitution. The conformance checklist replaces the Solution instead of testing it.
4. Ontology-light review. Wording is polished while kind, relation, support, evidence, measurement, assurance, and neighbouring-pattern authority remain unstable.
5. Decorative SoTA. Sources are cited, but they do not change the Solution, conformance checks, boundaries, examples, or relations.
6. Apparatus bloat. A draft accumulates fields, manifests, gates, and companion files that increase author and reader cost without improving admissible use.
7. Coordinate Goodharting. A draft is optimized for the declared coordinates until it becomes harder to use, costlier to maintain, or less faithful to the practical value the coordinates were meant to protect.
8. Endless improvement. Authors keep polishing because there is no declared stop condition and no visible distinction between material improvement and cosmetic movement.

#Forces

#Solution - Pattern quality as a scoped Q-Bundle over declared characteristics

State the scoped FPF pattern-quality read as a PatternQualityQBundle, not as one score.

#Architectural position

E.21 is a local receiving pattern for scoped FPF pattern-quality claims. It specialises existing FPF architecture; it does not create a new quality-governance subsystem.

It defines how to read whether one authored FPF pattern version is good enough for one declared reader, use, and scope, why improvement may stop, and which weaknesses remain bounded.

E.21 governs only these questions:

1. Which exact FPF pattern version is being read?
2. For which reader, use, scope, and currentness window?
3. Which hard blockers make comparison meaningless?
4. Which ordinal pattern-quality coordinates are active?
5. Which content evidence supports the coordinate readings?
6. Which PatternQualityStatus follows?
7. Why may improvement stop, or why must the claim narrow, repair, hold for architecture, or refresh?

E.21 does not govern:

• authoring the pattern body (E.8);
• running admission or refresh review profiles (E.19);
• general measurement legality (C.16, A.17, A.18, A.19);
• arbitrary engineering quality-family bundling (C.25);
• durable naming and local-first unification (F.18, E.10, A.6.P);
• project evidence, assurance, gate, work, release, safety, security, or compliance claims (A.10, B.3, A.20, A.21, A.15, or the exact receiving pattern).

A conforming E.21 read may cite neighbouring patterns, but it may not absorb their governed objects.

E.21 protects these pattern-quality boundary loci as part of the pattern-quality read: self-application, first-entry discoverability, publication and projection, bounded non-use, no forced winner, falsifiability and lowering evidence, reviewer-power misuse, AI, RAG, thin-echo summaries, and corpus ecology. A read that crosses one of these loci must keep the pattern-quality claim scoped, replayable, falsifiable, and kept under the exact neighbouring pattern application instead of becoming control apparatus.

#Mint/reuse and kind settlement

E.21 mints no new Kernel kind, no project-side evidence kind, no assurance kind, no work kind, no gate kind, no release kind, and no general maturity kind.

Its durable heads are local specialisations, fields, value sets, or scoped support constructs over existing FPF kinds:

Coordinate heads in E.21:4.3 are local pattern-quality characteristic heads inside PatternQualityEvaluationCharacteristicSpace. They do not become general FPF characteristics, numeric measures, maturity dimensions, or measurement templates unless a neighbouring C.16, A.17, A.18, or A.19 declaration makes that live.

A conforming E.21 read treats these heads as exact local constructs. When a head is reused outside E.21, the receiving text must either cite E.21 by value or rerun the exact governing pattern for the new use.

#Layered use architecture

E.21 has four activation layers.

Layer 0 and Layer 1 are ordinary; Layer 2 and Layer 3 are live-claim only.

Layer 0 - first-pass slice. Used for one first read of one pattern version. It names PatternVersionRef, WorkingReaderScope, IntendedUse, QualificationWindow, first admissible move evidence, activated blockers, minimal active coordinates, status, and next admissible repair or bounded non-use.

Layer 1 - content-evidence coordinate read. Used when the first-pass slice survives or when a substantive quality claim is being made. It checks activated EligibilitySet rows, active ordinal coordinates, and local CoordinateEvidenceRefs.

Layer 2 - comparison and stop. Used when several candidate edits or variants are live, or when a stop decision is being claimed. It activates DominanceSet, floors, TieBreakerSet, PatternQualityStatus, and StopCondition.

Layer 3 - optional refresh and trade-off support. Used only when live reuse, repeated findings, retrieval failure, high-use patterns, high-assurance reuse, contested neighbour authority, or variant-history retention makes support necessary. It may activate TelemetrySet, PatternQualityFront, PatternImprovementArchive, and full support cards.

A higher layer SHALL NOT be required merely because a lower-layer read exists.

Every activated layer must state what admissible use it buys.

If a field, card, telemetry signal, archive entry, or evidence record does not change first-pass usability, coordinate evidence, variant comparison, stop or reopen condition, or bounded non-use, it is apparatus bloat for the current claim.

#Self-application and recursion boundary

E.21 may be applied to itself only through the same lowest sufficient activation layer used for any other FPF pattern.

The ordinary self-application read closes when the current E.21 text exposes:

1. one first admissible action-guiding move for an author or reviewer;
2. the governed object of the quality read;
3. the no-single-score and no-administrative-proxy boundary;
4. the neighbour-authority boundary to E.8, E.19, C.25, C.16, A.17, A.18, A.19, F.18, E.10, A.6.P, E.17, A.10, B.3, A.20, A.21, and A.15;
5. one explicit stop or repair condition for the declared self-read scope.

Self-application SHALL NOT require an additional PatternQualityQBundle evaluating the previous PatternQualityQBundle. If the first self-read exposes a content defect, repair that defect in the pattern text or narrow the declared use. If the defect is architectural, use holdForArchitectureDecision.

#Pattern-quality question framing

A pattern-quality read does not begin from an implicit request such as "review this", "evaluate this", or "improve this". For a nontrivial read, use E.22 to declare the quality-read purpose before E.21 assigns pattern-quality coordinates or stop result.

The question frame states the QualityReadPurposeSelection, declared floor, desired improvement aim, protected trade-offs, and open-question classification rule. If no purpose is declared, the default is floorRead under the declared or receiving-pattern floor, not exceptionalImprovementRead.

exceptionalImprovementRead asks for non-dominated edits that could move active coordinates toward 5 where feasible. paretoTradeoffRead asks whether that movement damages usability, affordability, repair locality, corpus ecology, neighbour fit, entry and projection integrity, or another protected quality. openQuestionDiscoveryRead classifies important unasked questions as an existing E.21 coordinate issue, candidate coordinate or overlay, or outside E.21 under another exact FPF pattern. absorptionRead maps returned findings to coordinate movement, floor-only closure, unchanged already-satisfied content, trade-offs, quality loss, or outside-pattern disposition.

The question frame is not pattern-quality evidence by itself. It only states which E.21 question is being asked, so the resulting PatternQualityQBundle can answer the intended question rather than a reviewer-selected default.

When the same pattern version is being improved through repeated passes, use E.23 for the repeated quality-improvement method. E.21 supplies the pattern-quality coordinates, values, protected trade-offs, status, and stop meanings; it does not govern row-atomic absorption across passes, method-family selection, or stop, narrow, continue, switch method, or hold decisions.

#PatternQualityQBundle

PatternQualityQBundle := <PatternVersionRef, ClaimScope, WorkingReaderScope, IntendedUse, QualificationWindow, QualityReadQuestionFrameRef?, EligibilitySet, DominanceSet, CoordinateEvidenceRefs?, TieBreakerSet?, TelemetrySet?, EvidenceRefs, PatternQualityStatus, StopCondition> PatternQualityQBundle is the publication unit for a scoped pattern-quality claim.

PatternQualityQBundle is replayable when another reader can recover the same pattern version, declared scope, active eligibility rows, active coordinates, coordinate evidence refs, status payload, and stop/non-stop reason without chat memory, steward memory, or administrative placement state.

QualityReadQuestionFrameRef? is optional. When present, it cites an E.22 question frame that states whether this is a floor, exceptional-improvement, Pareto trade-off, open-question discovery, absorption, or combined read. The frame does not supply coordinate evidence; it constrains which E.21 result shape is expected.

Its first-pass slice may contain only the fields needed to decide whether one working reader can recover the governed object and first admissible move.

The full bundle is used when coordinate evidence, variant comparison, admission/refresh closure, high-assurance reuse, or contested neighbour authority is live.

The fields below are slots in that publication unit; they are not independent governed objects. The governed object remains the scoped pattern-quality claim for one pattern version.

PatternQualityStatus has this value set:

A PatternQualityStatus is an admissible-use posture for the pattern-quality claim. It is not a gate decision, release state, assurance level, compliance verdict, safety certificate, work authority, publication truth, or project-side refusal/approval.

A status without payload is not checkable. Every PatternQualityStatus SHALL carry the minimal status payload:

• admissibleForDeclaredUse: declared use, active floors, and remaining bounded non-use, if any.
• admissibleWithNarrowerUse: the exact narrowed ClaimScope, WorkingReaderScope, or IntendedUse.
• repairBeforeUse: the activated blocker and the first admissible repair move.
• holdForArchitectureDecision: the exact unresolved governed-object, neighbour-authority, split, or placement question.
• refreshNeeded: the exact SoTA, neighbour, terminology, retrieval, telemetry, or use-scope change that invalidated the previous read.

A bounded non-use result is a valid pattern-quality outcome.

When a pattern is not admissible for the declared ordinary use but remains useful as expert-only support, source-basis support, high-assurance support, historical rationale, narrow worked-case support, or neighbouring-pattern support, the PatternQualityStatus SHALL be admissibleWithNarrowerUse, not repairBeforeUse.

The narrowed use must name:

• the exact narrowed reader/use/scope;
• the prohibited broader use;
• the first admissible next pattern or repair if broader use is still desired.

#Coordinate readings are content readings, not administrative-state readings

PatternQualityEvaluationCharacteristicSpace coordinates measure the current pattern version and the content evidence available for that version: its recognition text, governed object, ontology, names, Solution, checklist, worked cases, SoTA rows, relations, and support boundaries. They do not measure whether the pattern has already been externally reviewed, merged into a monolith, included in a release branch, or accepted by one steward process.

Administrative state can change a PatternQualityQBundle only in these ways:

If a quality read says "low because this has not been externally reviewed" or "high because this is already in the monolith", it is using administrative state as a proxy. The repair is to state the substantive coordinate evidence directly and keep administrative state in ClaimScope, QualificationWindow, or receiving release/review patterns.

#Pattern property evidence vs reputation signals

Coordinate values read properties of the pattern version for the declared reader, use, scope, and qualification window. Popularity, adoption, award, steward praise, reviewer praise, external-review completion, number of reviews, monolith placement, release inclusion, or prior use is not a coordinate value and does not raise or lower a coordinate by itself.

Such signals may only serve as pointers to inspect content evidence. A replayable observation such as "three first-time readers could not recover the first admissible move from this Problem frame" may lower ActionPathGuidance because it names the pattern text, reader scope, use situation, observed failure, and coordinate affected. A signal such as "widely used", "not yet used", "reviewer-clean", "not externally reviewed", "popular", or "already accepted" remains outside the coordinate value unless it is rewritten into exact pattern-content evidence.

Absence of prior use, external review, landing, release, or steward acceptance is not evidence against 4 or 5. Only missing or weak pattern-content evidence for the declared use can lower the coordinate. The same pattern text under the same PatternVersionRef, ClaimScope, WorkingReaderScope, IntendedUse, and QualificationWindow should receive the same coordinate value whether it is still in a host file, already in the monolith, praised by reviewers, ignored by readers, or newly written.

#EligibilitySet: hard conditions before quality comparison

The EligibilitySet is not a low-scoring region. It is the precondition for meaningful comparison.

Eligibility rows have activation. A first-pass pattern-quality read always checks only the rows needed to decide whether one working reader can recognise the situation, recover the governed object, find the first admissible move, and avoid immediate neighbour-authority or apparatus overread.

Rows whose condition depends on a live source, live formal lens, live measurement claim, live accepted basis, live release/review state, live mission/pillar conflict, or live durable-name change activate only when that claim force is present in the pattern version or in the declared ClaimScope.

Failure of an activated eligibility row is a content blocker. A non-activated row is not a pass, not a waiver, and not a hidden todo; it is outside the current pattern-quality claim.

If an eligibility condition fails, the status is repairBeforeUse or holdForArchitectureDecision. Do not average the failure into better coordinates. A missing first move is not a weak coordinate. For an ordinary-use pattern-quality read, failed activated firstMoveRecoverability is a blocker unless the claim is explicitly narrowed to expert-only use, reference-only use, or architecture-decision use.

#PatternQualityEvaluationCharacteristicSpace

PatternQualityEvaluationCharacteristicSpace is the declared characteristic space for FPF pattern-quality reads. It uses ordinal coordinates. The default scale is a zero-based six-level scale:

A PatternQualityEvaluationCharacteristicSpace is not a general FPF quality ontology. Its coordinate heads remain local pattern-quality characteristic heads unless a neighbouring C.16/A.17/A.18/A.19 declaration makes a broader characteristic or measurement claim live.

A coordinate value in PatternQualityEvaluationCharacteristicSpace is an ordinal pattern-quality reading, not a U.Measure by default. It becomes a measurement claim only when the pattern explicitly declares a C.16 measurement template, scale, unit, or admissible coordinate construction. Otherwise the value is an evidence-backed ordinal judgement over pattern content for the declared scope.

The scale is zero-based because true absence is not a weak positive value. It uses six levels rather than ten because the read is ordinal: six levels distinguish absence, mere naming, partial expression, sufficiency, strong expression, and exceptional expression without pretending to have decimal-grade precision. The labels are intentionally domain-neutral. They describe degree of expression of whichever characteristic is being read; they do not import a substantive property such as robustness, stability, safety, maturity, completeness, usability, affordability, or evidence strength into every coordinate.

Authors may use a coordinate-specific named scale when needed, but they must keep the scale ordinal unless a C.16 measurement basis with the needed measurement semantics is declared. No arithmetic mean, percentage score, or hidden normalization is admissible.

The ordinal value of a coordinate is a content reading. FormalClaimLegalityAndLensFit = 3 means the formal or lens claim is sufficiently expressed for the declared use; it does not mean "not yet reviewed". FormalClaimLegalityAndLensFit = 5 means the same coordinate is exceptionally expressed in the current pattern text; it does not mean "already landed". The same pattern text in an extracted host and in a monolith should receive the same coordinate value unless the move changes the text, exposes a new content defect, changes the version under PatternVersionRef, or changes the declared use being evaluated.

Coordinate names in this section are local characteristic heads inside PatternQualityEvaluationCharacteristicSpace. They are not U.Measures and not general-purpose CHR patterns by name alone.

A coordinate value is an ordinal content reading over the pattern version text for the declared ClaimScope. It becomes a measurement claim only when a neighbouring C.16/A.17/A.18/A.19 declaration explicitly supplies the measurement basis, scale, unit, comparability mode, and evidence support.

Coordinate values in PatternQualityEvaluationCharacteristicSpace are ordinal content readings unless a neighbouring C.16/A.17/A.18/A.19 declaration makes a measurement claim live. DominanceSet compares these readings without scalarizing them.

#Coordinate value evidence test

A coordinate value is justified by content evidence, not by the label alone. The ordinary 4 wellExpressedForDeclaredUse test is:

1. the coordinate names the exact property being read;
2. the pattern text contains direct evidence for that property;
3. at least one positive case and one boundary or anti-case exercise the property;
4. neighbouring-pattern relations or non-use boundaries protect the property from overread;
5. SoTA or internal FPF architecture changes at least one Solution, checklist, relation, or worked-case line when the coordinate depends on a source or modeling lens;
6. the coordinate evidence does not depend on review completion, landing state, monolith placement, release state, or steward acceptance.

A 5 exceptionallyExpressedForDeclaredUse value requires the 4 test plus additional content evidence: multiple reinforcing loci, heterogeneous cases or anti-cases where the characteristic changes the result, explicit non-use boundary, and no hidden affordability, maintenance, neighbour-ripple, corpus, entry/projection, or proxy-for-value loss.

3 sufficientlyExpressedForDeclaredUse means the coordinate is usable for the declared scope but lacks one or more evidence references, cases, anti-cases, or boundary statements required for 4 or 5. Coordinate value and CoordinateEvidenceRefs remain distinct: a value says the declared expression degree for the characteristic; CoordinateEvidenceRefs say why that reading is justified.

The coordinates below are not one flat always-on audit grid. PatternQualityEvaluationCharacteristicSpace is an activation-normalized characteristic menu for scoped FPF pattern-quality reads.

A coordinate is admissible only when it reads a content property of the pattern version under quality read under the declared ClaimScope, WorkingReaderScope, IntendedUse, and QualificationWindow.

Hard blockers stay in EligibilitySet. Evidence kinds justify coordinate readings but are not coordinates by themselves. Telemetry reopens or calibrates a read but does not replace content evidence. Fronts and archives preserve candidate trade-offs but do not add ordinary drafting obligations. PatternQualityStatus and StopCondition are results of the read, not coordinates. Project-side evidence, assurance, gate, release, work, safety, security, or compliance claims stay under exact neighbouring patterns.

Each coordinate has an activation class:

• first-pass core: ordinarily active for the first ordinary read of one pattern version;
• ordinary stop core: active when the read claims admission, stop, repair-before-use, or narrowed-use closure;
• claim-support: active when evidence, currentness, SoTA, case breadth, measurement, formal lens, replay, or high-value claim support is live;
• corpus/publication: active when the edit changes entry cues, ToC/J.4/Preface, summaries, cards, dashboards, retrieval, durable names, relations, projections, or neighbouring authority;
• front/refresh: active when variants, non-dominated fronts, archives, refresh, repeated findings, or no-forced-winner cases are live.

Inactive coordinates are not passes, waivers, or hidden failures. They are outside the current pattern-quality claim.

Activated overlays. These overlays are not ordinary coordinates. They become active only when the pattern version or declared ClaimScope makes the issue live.

#DominanceSet, TieBreakerSet, TelemetrySet

DominanceSet names the subset of activated characteristics selected for the current ClaimScope. The coordinate menu in E.21:4.3 is not one always-on grid.

For one first ordinary read of one pattern version under quality read, the minimal active coordinates are:

• WorkingSituationAndUseBoundaryRecognizability;
• ActionPathGuidance;
• GovernedObjectAndClaimScopeStability;
• NeighborAuthorityAndBoundedUseFit;
• PracticalUseDeltaAndHarmPrevention;
• UseAffordabilityAndApparatusProportionality.

For admission, stop, repair-before-use, or narrowed-use closure, add the ordinary stop core coordinates:

• ClosureAndBoundedNonUseRecoverability;
• SemanticKindAndNameRecoverability when names/kinds carry FPF force beyond eligibility minimum;
• RepairLocalityAndChangeImpactPredictability when an edit or repair is being judged;
• ProxyForValueSubstitutionResistance before any stop or comparison claim.

Activate claim-support, corpus/publication, front/refresh coordinates, and activated overlays only when the pattern version, candidate edit, or declared ClaimScope makes their claim force live.

Inactive coordinates are outside the current pattern-quality claim. They are not passes, waivers, or hidden failures.

A candidate pattern version A dominates candidate B only when:

1. both pass the active EligibilitySet;
2. A is no worse than B on every active dominance coordinate;
3. A is better on at least one active dominance coordinate; and
4. A does not create a new hard blocker, an unacceptable drop in UseAffordabilityAndApparatusProportionality, RepairLocalityAndChangeImpactPredictability, or ProxyForValueSubstitutionResistance, or an unacceptable unscored increase in reader, author, maintainer, migration, evidence, entry/projection, corpus, or neighbour-integration cost.

TieBreakerSet is used only among non-dominated candidates. Default tie-breakers are:

Cost tie-breakers are not a hiding place for live quality loss. If reader, author, reviewer, maintainer, migration, evidence, entry/projection, corpus, or neighbour-integration cost can change admissible use, represent it through UseAffordabilityAndApparatusProportionality, RepairLocalityAndChangeImpactPredictability, ExternalEntryAndProjectionIntegrity, PatternLanguageEcologyFit, ProxyForValueSubstitutionResistance, or another active coordinate before using tie-breakers.

Support material retention test. A support card, telemetry fixture, archive row, worked slice, proof sketch, or companion note remains attached to the pattern-quality read only when the read states what real quality breakage would return if it were absent.

If no breakage is named, fold the useful content into the pattern body, keep it as accepted basis only, or remove it from the active quality read.

High-assurance separation rule. When high-assurance reuse needs additional evidence, proof sketches, telemetry, or support cards, preserve the ordinary pattern body's first admissible move unless the ordinary use itself changes.

High-assurance material should normally live in a named support card, worked slice, or neighbouring evidence/assurance pattern. It should not be inserted into the ordinary Solution if doing so makes first use harder for the declared ordinary reader.

TelemetrySet is optional for early drafts and useful for mature or high-use patterns. Telemetry is an activation signal for reopening or calibration, not a standing requirement for every pattern-quality read. Typical telemetry includes:

Refresh opens the smallest live locus: the affected source stance, neighbour relation, coordinate reading, worked case, name, eligibility row, or status payload. It reopens the whole pattern-quality read only when that local change can change PatternQualityStatus or StopCondition.

#StopCondition

Improvement can stop for the declared scope only when the PatternQualityQBundle satisfies:

StopCondition :=
  EligibilitySet passes
  AND the pattern text is an action-guiding FPF pattern under E.8 for the declared use
  AND the first move is recoverable from Problem frame and Solution, or the claim is narrowed to support-only use
  AND no open P1/P2 pattern-quality findings for the declared ClaimScope
  AND all active DominanceSet coordinates meet the declared floor
  AND no active coordinate is 0 or 1 for a use the pattern claims to support
  AND the current candidate is non-dominated on the active DominanceSet
  AND the next proposed edits are TieBreaker-only or cosmetic for this scope
  AND no active improvement creates unmeasured usability, affordability, repair-impact, entry/projection, corpus-ecology, or neighbour-ripple loss outside the DominanceSet
  AND remaining weaknesses are expressed as bounded non-use or a named receiving pattern
  AND TelemetrySet, when active, shows no recurring blocker class

No stop condition may close while a visible coordinate improvement creates unmeasured loss in first-use cost, reader comprehension, maintainer locality, neighbour stability, bounded non-use clarity, or practical payoff.

The ordinary floor for an admission-ready pattern is 3 sufficientlyExpressedForDeclaredUse on every active coordinate. A lower coordinate may remain only by narrowing ClaimScope, WorkingReaderScope, or IntendedUse; it is not hidden in an average.

#PatternQualityFront and PatternImprovementArchive

A PatternQualityFront is live only when at least two candidate versions or candidate edits are being compared.

A PatternImprovementArchive is live only when preserving rejected or near-miss variants changes future repair, refresh, or selection.

Do not create either construct for a single ordinary repair unless a candidate comparison or contested stop decision is already live.

When these constructs are live, keep:

This is the pattern-quality analogue of NQD/front discipline. It prevents "one winner" thinking while still allowing one candidate to be selected for the current ClaimScope.

The archive is not required for every small edit. It becomes useful when several plausible versions trade off SemanticKindAndNameRecoverability, reader cost, SoTABindingAndCurrentness, CaseCountercaseAndTransferCoverage, UseAffordabilityAndApparatusProportionality, RepairLocalityAndChangeImpactPredictability, or ProxyForValueSubstitutionResistance.

PatternImprovementArchive records candidate-edit trade-off evidence only. It is not chat memory, process chronology, review history, backlog, or permanent appendix.

#No forced winner

When several candidates are non-dominated under the active EligibilitySet, DominanceSet, and cost coordinates, E.21 SHALL NOT force one winner unless the declared IntendedUse requires one selected candidate.

If ordinary drafting, exploration, or architecture discussion remains live, the admissible output may be a scoped PatternQualityFront with each candidate's bounded use and known trade-off. A single selected version is required only when admission, landing-support, release-support, high-assurance reuse, or another receiving action requires one candidate by value.

Choosing one candidate from a non-dominated front requires either:

• a declared TieBreakerSet that does not override hard blockers or live quality loss;
• a narrowed ClaimScope or WorkingReaderScope;
• a named neighbouring decision pattern when the choice is no longer a pattern-quality read.

#Relationship to E.19

E.21 supplies the characteristic space and stop condition for pattern-quality claims. E.19 supplies the review profile, findings-first run record, and admission/refresh outcome.

Use them together this way:

1. Use E.8 to author the pattern.
2. Use E.21 to state the quality bundle and non-scalar stop condition.
3. Use E.19 to run the selected review profiles and record findings.
4. If an E.19 finding changes the quality read, update the PatternQualityQBundle; do not turn the E.19 verdict into a scalar score.
5. If the pattern version is being improved through repeated passes, use E.23 for the repeated method while E.21 continues to supply the pattern-quality values and stop meanings.

An E.19 pass, return, or absence is not itself a coordinate value. It may supply evidence about the pattern text, expose a content defect, justify a confidence judgement, or constrain the admissible ClaimScope; it does not change the same FormalClaimLegalityAndLensFit, SemanticKindAndNameRecoverability, ActionPathGuidance, or ClosureAndBoundedNonUseRecoverability coordinate reading by administrative state alone.

Do not call an E.21 EligibilitySet, PatternQualityStatus, StopCondition, or coordinate floor a gate. E.21 states a scoped pattern-quality read. E.19 runs pattern-quality review profiles. A.21 governs operational gate decisions. These three uses do not collapse by metaphor.

An E.21 result is still a pattern-quality result. It is not project evidence, safety certification, gate passage, assurance acceptance, work authority, release approval, or publication truth unless the exact receiving pattern opens that project-side relation.

#Minimal PatternQualityQBundle card

Ordinary first-pass slice. This is not a new kind. It is the smallest admissible slice of PatternQualityQBundle for one first read of one pattern version under quality read.

PatternQualityQBundle / first-pass slice:
  PatternVersionRef:
  WorkingReaderScope:
  IntendedUse:
  QualificationWindow:
  FirstMoveEvidenceRef: <Problem frame / Solution / named neighbouring-pattern application>
  EntryBlockers: <none | first move absent | governed object drift | checklist-as-solution | apparatus overreach | neighbour overread | other activated blocker>
  MinimalDominanceSet: <default six coordinates, plus activated coordinates if live>
  PatternQualityStatus:
  Next admissible repair or bounded non-use:

The full one-screen card is used when the first-pass slice survives or when the declared use requires coordinate evidence, variant comparison, admission, refresh, high-assurance reuse, or contested stop.

The fuller card remains one screen when Layer 1, comparison, admission, refresh, high-assurance reuse, or contested stop makes it live:

PatternQualityQBundle:
  PatternVersionRef: <pattern id + edition or host path>
  ClaimScope: <ordinary authoring support quality read | admission-support quality read | refresh-support quality read | landing-support quality read | external-review-support quality read | high-assurance-reuse support quality read | ...>
  WorkingReaderScope: <primary reader + first-use situation>
  IntendedUse: <what this quality read may support>
  QualificationWindow: <edition / SoTA / neighbour / release window>
  EligibilitySet: <pass/fail rows with blockers>
  DominanceSet: <selected coordinates + floor>
  CoordinateEvidenceRefs: <text sections, worked cases, SoTA rows, relation checks, or findings that support coordinate values; not placement/review state alone>
  PatternQualityStatus: <value set from E.21:4.1>
  StopCondition: <satisfied | not satisfied, with exact reason>
  EvidenceRefs: <worked cases, SoTA rows, review refs, source refs>

#Publication and projection boundary for quality cards

A rendered PatternQualityQBundle, first-pass slice, quality table, status badge, or dashboard tile is a publication or projection of the pattern-quality read. It is not the pattern version, not the authority source, not project evidence, not release approval, not gate passage, and not assurance acceptance.

When the rendered card is used as a bounded publication unit, the card SHALL keep visible:

1. the exact PatternVersionRef;
2. the carried move: scoped pattern-quality read, first-pass slice, variant comparison, stop/non-stop reason, or bounded non-use;
3. the outside boundary to project evidence, assurance, gate, release, work, and publication truth;
4. the exact receiving pattern when a downstream claim is live.

A ToC row, J.4 row, README note, dashboard tile, or generated summary may echo an E.21 result only as a thin orientation cue unless it cites the full governed quality read by value.

Generated summaries, retrieval snippets, README lines, ToC reminders, and J.4 entries may expose an E.21 result only as thin echo or controlled coarsening.

A thin echo may say:

• <PatternVersionRef> has PatternQualityStatus = repairBeforeUse for ordinary use because first move is absent. See <quality read ref>.

A thin echo SHALL NOT say:

• approved pattern;
• safe pattern;
• compliant pattern;
• quality passed;
• do not use this pattern without scope;
• E.21 certified this pattern.

If a generated or coarsened rendering is used for reliance beyond orientation, apply the exact receiving pattern live: A.6.3.CSC, E.17.EFP, E.17.AUD, A.10, or B.3, as applicable.

#Reader move loop

The fast entry loop for one pattern version under quality read is:

1. write the pattern-version line: <PatternVersionRef> for <WorkingReaderScope> under <IntendedUse> within <QualificationWindow>;
2. read only the pattern version's Problem frame and Solution until one first admissible action-guiding move is recoverable;
3. if no first move is recoverable, or if the move lives only in the Conformance Checklist, assign PatternQualityStatus = repairBeforeUse for the declared use unless a narrower support-only use is explicitly named;
4. if the first move is recoverable, check the default six first-pass coordinates and any activated coordinates;
5. expand to the full comparison and stop loop only when a stop decision, variant comparison, admission, refresh, high-assurance reuse, or contested neighbour claim is live.

The full comparison and stop loop is:

1. name the exact PatternVersionRef;
2. declare ClaimScope, WorkingReaderScope, IntendedUse, and QualificationWindow;
3. check activated EligibilitySet rows before comparing coordinates;
4. read each active coordinate from CoordinateEvidenceRefs, not from administrative state;
5. ask what became worse when the visible coordinates improved: first-use affordability, author/reviewer effort, repair-impact predictability, entry/projection integrity, corpus ecology, neighbour ripple, and proxy-for-value substitution;
6. compare candidates through DominanceSet only after eligibility passes and Goodhart-risk questions are visible;
7. use TieBreakerSet only among non-dominated candidates;
8. assign one PatternQualityStatus;
9. state whether StopCondition is satisfied or which bounded non-use, receiving pattern, or content repair remains.

If the reader cannot recover the fast-entry result for one pattern version, activated firstMoveRecoverability fails for ordinary use even when the prose is polished. If the first move is recoverable but the reader cannot continue the admissible action path, ActionPathGuidance is below 4 wellExpressedForDeclaredUse for ordinary use. If the reader cannot perform the closure and bounded non-use loop when that loop is live, ClosureAndBoundedNonUseRecoverability cannot support stop closure for that declared scope.

#CoordinateEvidenceRefs support card

CoordinateEvidenceRefs may be inline section references, short claims, review findings, worked slices, or full support cards. Full support cards are required only when the coordinate value is contested, reused for high-assurance closure, used across several candidate variants, or cited outside the local pattern-quality read.

CoordinateEvidenceRefs is the local evidence list for coordinate values. It is support for coordinate readings, not a coordinate family by itself.

CoordinateEvidenceRef := <Coordinate, EvidenceKind, HostSectionRef, Claim, Limitation, LoweringCondition?>

LoweringCondition? is required only when a coordinate value is 4 or 5, or when the coordinate supports admissibleForDeclaredUse or StopCondition.

It states one concrete content discovery, reader failure, neighbour conflict, SoTA change, worked-case counterexample, cost increase, entry/projection overread, corpus-ecology conflict, or affordability loss that would lower the coordinate or reopen the read.

A lowering condition is not a test suite and not a review plan. It is a falsifiability hook for the current pattern-quality claim.

Default evidence kinds:

entryProjectionEvidence and corpusEcologyEvidence are active only when the candidate edit changes durable names, relations, ToC/J.4/Preface entry support, retrieval-facing cues, support projections, summaries, cards, dashboards, or neighbouring-pattern authority. They are not required for local wording repair.

Example support card:

CoordinateEvidenceRef:
  Coordinate: FormalClaimLegalityAndLensFit
  EvidenceKind: measurementScaleEvidence + workedCaseEvidence + sotaImplicationEvidence
  HostSectionRef: E.21:4.3, E.21:4.12, E.21:11
  Claim: Q-Bundle + ordinal characteristic space + Pareto/front lens preserve the comparisons needed for pattern improvement while rejecting scalar averages.
  Limitation: Not a statistical estimator, project assurance result, or tool mandate.
  LoweringCondition: If an E.21 edit permits arithmetic averaging of ordinal coordinate values, lower this coordinate to <=2 for ordinary use.

#Pattern-quality finding sentence grammar

A pattern-quality finding is admissible only when it has this shape:

Pattern-quality finding:
  PatternVersionRef: <PatternVersionRef>
  ClaimScope / reader / use / window:
  Finding kind: <eligibility blocker | coordinate reading | status payload | stop-condition failure | bounded non-use | neighbouring-pattern application>
  Exact E.21 locus: <EligibilitySet row | coordinate | status | stop clause>
  Content evidence: <HostSectionRef / worked case / SoTA row / relation check / review finding>
  Result: <PatternQualityStatus effect or coordinate value effect>
  First admissible repair or bounded non-use:

Forbidden finding shapes:

• weak pattern;
• not FPF enough;
• quality low;
• review failed;
• needs more evidence;
• not safe/compliant;
• too complex;
• not ready.

These are admissible only after rewriting into the exact [E.21](/generated/patterns/E.21) locus, content evidence, status effect, and first admissible repair or bounded non-use.

#Local name-precision cards for E.21 heads

These cards are local name-precision evidence for E.21; they are not a separate glossary. Coordinate names are defined in E.21:4.3 and are not repeated here unless the token also names a local construct.

The E.21:4.11 local name-precision cards are the local F.18-compatible settlement for E.21 heads. They do not require separate full Name Card artifacts for ordinary use.

A full F.18 Name Card is required only when an E.21 head is reused outside E.21, collides with an existing FPF head, enters a UTS/Concept-Set row, or becomes a durable cross-pattern naming decision rather than a local field/value-set name.

Plain twins for ordinary reading. These Plain twins are reader aids only; the Tech head remains authoritative.

#Mathematical lens adequacy proof sketch

The mathematical lens in E.21 is a finite ordinal characteristic-space lens plus a Pareto/front comparison over candidate pattern versions.

The lens is admissible because pattern quality is multi-characteristic, ordinal, and scope-dependent. It is not admissible for estimating user adoption, certifying project safety, proving product compliance, or assigning a universal quality number. The stop condition closes the lens: once eligibility passes, floors are met, the candidate is non-dominated, remaining weaknesses are bounded, no active coordinate improvement hides affordability, repair-impact, entry/projection, corpus-ecology, or neighbour loss, and live telemetry shows no recurring blocker class, additional edits must show a real front movement rather than a cosmetic preference.

#Archetypal Grounding

Tell. A good FPF pattern is not merely complete, attractive, or formally strict. It is an action-guiding method description whose quality is a scoped bundle: it helps the right reader recognise the right situation, make the next admissible move, avoid the wrong neighbour, and trust the result for the declared use.

Recognition matrix.

Show - System pattern. A system-architecture pattern has precise structure vocabulary and a current, exact SoTA row, but its opening never says what an engineer-manager should do first. Activated firstMoveRecoverability fails for ordinary use. The admissible result is not "quality 78/100"; it is repairBeforeUse for admission, or admissibleWithNarrowerUse if it is kept as expert-only support.

Show - First-pass pattern-version read. An author opens a candidate FPF pattern and writes: PatternVersionRef = C.xx@draft-3; WorkingReaderScope = engineer-manager using the pattern for first ordinary application; IntendedUse = continue drafting or repair before use; QualificationWindow = current FPF edition.

The author reads only Problem frame and Solution. If the pattern says which object it governs but gives no first admissible action-guiding move, the first-pass read closes: activated firstMoveRecoverability fails and PatternQualityStatus = repairBeforeUse for ordinary use. No PatternQualityFront, TelemetrySet, or full coordinate table is needed.

If the first move exists but the continuation path or apparatus required to apply it is heavier than the ordinary case can justify, activate ActionPathGuidance and UseAffordabilityAndApparatusProportionality. The repair is to keep the ordinary first move light and move heavier support to a named neighbouring pattern or high-assurance support card.

Show - Episteme pattern. A publication or evidence pattern reads beautifully and has many examples, but it treats an E.19 pattern-quality result as project assurance. NeighborAuthorityAndBoundedUseFit and ClaimSupportTraceabilityCurrentnessAndReplayability fail because assurance belongs in B.3 and evidence/currentness belongs in A.10. If a dashboard tile or generated summary repeats the result without scope and status payload, ExternalEntryAndProjectionIntegrity also fails. The repair is to keep the pattern-quality result as pattern-quality evidence and open the exact project-side receiving relation only when needed.

Show - Mathematical lens pattern. A pattern says that it uses a characteristic space, but then compares variants by "overall quality". FormalClaimLegalityAndLensFit fails because the ordinal coordinates are being collapsed and the preserved and rejected structures are not named. The repair is to state the finite ordinal coordinate set, eligibility predicate, dominance relation, tie-breaker boundary, and non-admissible scalar uses.

Show - New pattern candidate. Two drafts of the same pattern both pass eligibility. Draft A is shorter and easier to read; Draft B has better SoTA, case/countercase breadth, and neighbour closure but adds a heavier bundle card. Neither dominates if Draft B's extra apparatus is live only for high-assurance reuse. The correct output may be one ordinary draft plus a support card for high-assurance use, not one averaged winner.

Show - Goodhart trade-off. An author raises visible E.21 coordinates to high values by adding proof sketches, support cards, and SoTA rows. The pattern now reads better on the visible table, but a cold author needs more time to find the first move and a maintainer must update more named sections or evidence records after each small repair. The quality read cannot stop until UseAffordabilityAndApparatusProportionality, RepairLocalityAndChangeImpactPredictability, and ProxyForValueSubstitutionResistance are read by content evidence. If those coordinates fall, the candidate is not an improvement for the declared ordinary-use scope; it may become a high-assurance variant instead.

Show - Self-application support card.

CoordinateEvidenceRef:
  Coordinate: ActionPathGuidance
  EvidenceKind: recognitionTextEvidence + workedCaseEvidence
  HostSectionRef: E.21:1, E.21:4.9, E.21:5
  Claim: The reader has a first move, a continuation loop, and heterogeneous examples that show when the loop changes the quality result.
  Limitation: This is pattern-quality guidance, not project-side work authorization.

#Bias-Annotation

E.21 intentionally biases evaluation away from single scores, reviewer taste, and flat audit grids and toward activation-normalized characteristic spaces. This supports P-1 Cognitive Elegance, P-2 Didactic Primacy, P-7 Pragmatic Utility, P-10 Open-Ended Evolution, and P-11 SoTA Alignment.

Lens cautions:

#Architectural characteristics preserved by this pattern

This table states which architectural characteristics E.21 protects; it does not create a separate review process.

#Conformance Checklist

| CC-E21-19 (First action before control). | A pattern-quality read SHALL recover the pattern version's first admissible action-guiding move from Problem frame and Solution before adding checklist, telemetry, archive, or high-assurance support. If the first move is absent or only appears in the checklist, the read may close as repairBeforeUse or admissibleWithNarrowerUse. | Prevents conformance checks and control apparatus from replacing pattern guidance. | | CC-E21-20 (SoTA mutation test). | Every live SoTA row SHALL state which E.21 field, eligibility condition, coordinate, worked slice, relation, conformance item, non-use boundary, or stop/reopen condition changes because of the adopted/adapted/rejected stance. A source that changes no content-bearing text is rationale support, not SoTA binding. | Prevents decorative SoTA and keeps SoTABindingAndCurrentness content-bearing. | | CC-E21-21 (SoTA currentness and lineage split). | A foundational, official, popular, or lineage source MAY remain in E.21:11, but a current-practice claim SHALL either cite a current SoTA anchor under E.8 or explicitly mark the older or official source as lineage-only, current-standard reference, rationale-only, or rejected-popular-practice material. | Prevents old standards, fresh standards, classic papers, and popular practice from masquerading as present SoTA. | | CC-E21-22 (Evaluation non-certification). | A pattern-quality evaluation SHALL NOT be used as safety, security, compliance, release, project assurance, or gate certification. When such a claim is live, the read SHALL open the exact receiving FPF pattern and state the supported and unsupported use. | Blocks audit-theatre and compliance-by-checklist overread. | | CC-E21-23 (Activated retrieval evidence only). | Retrieval, RAG, search, or misentry telemetry SHALL be used only when retrieval-facing pattern entry or observed misretrieval is live; it SHALL NOT become a universal benchmark requirement for ordinary pattern drafts. | Keeps modern retrieval evaluation useful without adding review bureaucracy. | | CC-E21-24 (First-pass content slice). | A conforming first-pass pattern-quality read SHALL be able to close on the smallest slice that identifies pattern version, reader/use/window, first admissible move evidence, activated blockers, minimal dominance coordinates, status, and next admissible repair or bounded non-use. | Keeps E.21 usable as content guidance rather than review bureaucracy. | | CC-E21-25 (Status payload). | Every PatternQualityStatus SHALL state the exact use, scope, reader boundary, blocker, reopen trigger, or architecture-decision question that makes the status true. | Prevents status labels from becoming vague maturity tags. | | CC-E21-26 (Ordinal-reading boundary). | A coordinate value SHALL be treated as an ordinal content reading unless a C.16 measurement basis is explicitly declared. | Prevents accidental pseudo-measurement. | | CC-E21-27 (Claim-triggered activation). | Coordinates and eligibility rows SHALL be activated by the live claim force of the pattern version or quality claim; inactive rows SHALL NOT be treated as hidden failures or waived passes. | Keeps the complete characteristic space from becoming a universal audit grid. | | CC-E21-28 (Kind settlement). | Every durable or FPF-force-bearing E.21 head SHALL be classified as an existing FPF kind specialisation, local field, value set, local evidence-reference record, or scoped support construct. A head with no recovered kind SHALL NOT be used in a stop decision. | Prevents E.21 from minting a parallel quality ontology. | | CC-E21-29 (No gate/status overread). | PatternQualityStatus, EligibilitySet, coordinate floors, and StopCondition SHALL NOT be described as gate passage, release status, role state, assurance level, work authority, or project approval. | Keeps E.21, E.19, A.21, B.3, and release/work patterns distinct. | | CC-E21-30 (Local name-card sufficiency). | E.21:4.11 local name-precision cards are sufficient for ordinary E.21 use. A full F.18 Name Card is required only when a head is reused outside E.21, collides with an existing head, enters durable cross-pattern vocabulary, or changes naming authority. | Preserves naming discipline without turning ordinary pattern-quality reading into naming bureaucracy. | | CC-E21-31 (Coordinate-head scope). | Coordinate heads in PatternQualityEvaluationCharacteristicSpace SHALL remain local ordinal characteristic heads for pattern-quality reads unless a neighbouring C.16/A.17/A.18/A.19 declaration promotes a specific coordinate into a measurement or broader characteristic claim. | Prevents accidental metrics, maturity dimensions, and pseudo-measurement. | | CC-E21-32 (Neighbour-governed claim boundary). | A conforming E.21 read SHALL state or preserve the governing-pattern boundary between E.21 and live neighbouring patterns when authoring, review, measurement, naming, evidence, assurance, gate, release, work, or project-side claims are involved. | Prevents E.21 from becoming a central quality-governance subsystem. | | CC-E21-33 (Layer activation). | A conforming E.21 read SHALL use the lowest activation layer sufficient for the declared claim. Front, archive, telemetry, and full support-card apparatus SHALL NOT be required for a first-pass read unless the claim makes them live. | Preserves affordability and prevents bureaucracy. | | CC-E21-34 (Replayable quality read). | A PatternQualityQBundle SHALL be replayable from its pinned pattern version, reader/use/scope/window, active eligibility rows, active coordinates, evidence refs, status payload, and stop/non-stop reason, without relying on chat memory or administrative placement state. | Preserves auditability without requiring a process log. | | CC-E21-35 (No neighbour substitution). | E.21 SHALL NOT absorb the governed object of E.8, E.19, C.25, C.16/A.17/A.18/A.19, F.18/E.10/A.6.P, or project-side evidence, assurance, gate, work, and release patterns. When such a claim is live, E.21 SHALL name the exact receiving pattern application by value. | Preserves modularity and composability. | | CC-E21-36 (Smallest live reopen). | A refresh or telemetry signal SHALL reopen the smallest affected locus: source stance, neighbour relation, coordinate reading, worked case, name, eligibility row, status payload, or stop condition. The whole quality read reopens only when that local change can change status or stop. | Preserves evolvability without whole-pattern churn. | | CC-E21-37 (Status is not authority). | PatternQualityStatus SHALL remain an admissible-use posture for the pattern-quality claim and SHALL NOT be used as project approval/refusal, gate decision, release state, assurance level, compliance verdict, safety certificate, or work authority. | Preserves scope safety and trust calibration. | | CC-E21-38 (No quality veto without content locus). | A blocking pattern-quality finding SHALL name the exact activated eligibility row, coordinate, status payload, or stop-condition clause, plus content evidence and the first admissible repair or bounded non-use. | Prevents pattern-quality review from becoming reviewer authority theatre. | | CC-E21-39 (Self-application closure). | E.21 self-application SHALL use the lowest sufficient activation layer and SHALL NOT require recursive quality bundles evaluating quality bundles. | Prevents infinite regress and keeps E.21 usable. | | CC-E21-40 (Thin echo boundary). | ToC rows, J.4 rows, README notes, dashboards, generated summaries, and retrieval snippets SHALL NOT replace the governed PatternQualityQBundle; they may only echo it by value and scope. | Prevents projection authority and RAG/snippet overread. | | CC-E21-41 (No forced winner). | When multiple candidates are non-dominated and no receiving action requires one selected candidate, a conforming read SHALL NOT force a single winner. | Preserves NQD/front discipline and blocks hidden scalarization. | | CC-E21-42 (Bounded non-use as valid outcome). | admissibleWithNarrowerUse SHALL be used when a pattern is not ordinary-use admissible but remains useful for a named narrower reader/use/scope. | Prevents unnecessary rewrite churn and preserves useful legacy/support material. | | CC-E21-43 (High-value falsifiability hook). | Coordinate values 4 or 5, and any coordinate supporting admissibleForDeclaredUse or StopCondition, SHALL state a lowering condition or content discovery that would reopen or lower the read. | Makes high quality claims falsifiable without adding a full harness. | | CC-E21-44 (Support retention test). | Support material SHALL remain active only when the read states what quality breakage would return if that material were absent. | Prevents support material from becoming folklore, hidden authority, or permanent reader cost. | | CC-E21-45 (Pattern text vs pattern application). | E.21 SHALL NOT be used to certify that a project correctly applied a pattern. It reads the quality of the pattern version; project/application claims remain under exact receiving patterns. | Preserves FPF-side and project-side boundaries. | | CC-E21-46 (High-assurance separation). | High-assurance support SHALL NOT make the ordinary pattern body harder to use unless the ordinary use itself changes. | Keeps ordinary action guidance alive while allowing additional support material where live. | | CC-E21-47 (Activation-normalized coordinates). | PatternQualityEvaluationCharacteristicSpace SHALL NOT be used as one flat always-on audit grid. Each coordinate SHALL state its activation class, and inactive coordinates SHALL NOT count as pass, waiver, or hidden failure. | Prevents characteristic bloat and hidden checklist control. | | CC-E21-48 (Hard blockers stay out of dominance). | firstMoveRecoverability, hard measurement illegality, shadow neighbour authority, administrative proxy use, and live mission/pillar conflict SHALL be treated as eligibility blockers when activated, not weak coordinate values. | Prevents hard failures from being averaged or front-compared away. | | CC-E21-49 (No evidence-as-coordinate substitution). | CoordinateEvidenceRefs, evidence kinds, support cards, review findings, and telemetry signals SHALL justify, reopen, or calibrate coordinate readings; they SHALL NOT become coordinates by themselves. | Keeps the quality space about pattern properties, not support artefact volume. | | CC-E21-50 (No hidden double weighting). | When two quality concerns are directly coupled, the pattern SHALL either merge them into one coordinate with explicit subreadings or explain why the two coordinates can fail independently and require different repairs. | Prevents the number of coordinate rows from acting like a hidden weighting scheme. | | CC-E21-51 (Formal-claim activation). | Measurement, score, comparison, threshold, aggregation, mathematical-lens, causal-lens, QL-lens, simulation, representation, or learned-lens checks SHALL activate FormalClaimLegalityAndLensFit; absence of such a claim SHALL NOT create an ordinary coordinate obligation. | Prevents ordinary pattern reads from becoming formal-method bureaucracy. | | CC-E21-52 (Projection and corpus activation). | External entry, publication projection, retrieval, RAG, dashboard, durable-name, relation, or corpus-ecology coordinates SHALL activate only when the pattern version or candidate edit changes those surfaces or is known to be misentered or overread through them. | Keeps corpus safety without universal projection bureaucracy. | | CC-E21-53 (High-value lowering condition). | Coordinate values 4 or 5, admissibleForDeclaredUse, and stop claims SHALL state a concrete lowering or reopen condition unless the declared use is only first-pass repair triage. | Makes high-value pattern-quality claims falsifiable without requiring a test harness. |

#Common Anti-Patterns and How to Avoid Them

#Consequences

#Rationale

FPF patterns are not essays, APIs, step lists, or generic best-practice notes. They are action-guiding method descriptions for recurring working situations. Their quality therefore cannot be read from style, section count, one score, or one flat audit grid.

C.25 already shows why composite quality claims need bundles. C.16, A.17, and A.18 show why measured values require declared characteristics and scales. A.19 shows why multi-coordinate reasoning needs a declared characteristic space. F.18 and NQD/front discipline show why several viable candidates can remain non-dominated. E.12 and E.13 show why usability, cognitive ergonomics, practical utility, and proxy-for-value substitution must remain quality-bearing rather than administrative afterthoughts. E.19 shows why review should use triggered profiles rather than one giant checklist.

The characteristic space is activation-normalized because FPF pattern quality is read from the live claim made by the pattern version. A first-pass ordinary-use read needs entry, first move, action path, bounded non-use, and a small evidence slice for the active quality claim; a formal-claim read also needs legality of measurement, comparison, mathematical lens, causal lens, quantum-like lens, simulation, representation, or learned model; a publication or retrieval-facing read also needs entry/projection integrity and corpus-ecology fit. Treating every coordinate as always active would turn E.21 into hidden checklist control.

The coordinate set also avoids hidden double weighting. Concerns that usually fail and repair together are merged with explicit subreadings; a weak live subreading limits the coordinate value instead of being averaged away. Concerns that fail independently remain separate because their repair moves differ: action path is not closure, semantic/name recovery is not neighbour authority, claim support is not project assurance, and local pattern quality is not corpus ecology.

E.21 applies existing FPF architecture to pattern quality. It gives authors a way to say: this version is admissible for this use, under these active coordinates, with these hard blockers cleared, and these remaining weaknesses bounded. That is more reviewable than "good enough" and lighter than pretending every pattern needs a universal maturity score.

#SoTA-Echoing

E.21:11 is a SoTA-binding table, not a bibliography. A row is live only when it changes at least one E.21 field, eligibility condition, coordinate, worked slice, relation, conformance item, non-use boundary, or stop/reopen condition, and it uses SoTA in the E.8 sense: current best-known problem-solving practice for the governed problem.

If a SoTA Synthesis Pack@CG-Frame exists for pattern-quality evaluation, this section cites its claim IDs and does not fork an untracked SoTA narrative. If no pack exists, this section is a provisional seed and must still state adopt | adapt | reject, the concrete E.21 effect, and the boundary of non-overread.

A source that only supplies lineage, popularity, or familiar terminology is not a SoTA row. It may remain as rationale material, but it does not satisfy SoTABindingMinimum.

#Relations

| E.9.DA | DRR decision-adequacy reads before downstream authoring use. | Open only when a pattern-quality defect traces to an upstream DRR whose selected answer, receiving-locus disposition, source-use carry-through, accepted-decision carry-through, or architecture selection is missing, vague, unassigned, source-theatre-like, or architecture-by-addressing. | Treat upstream DRR adequacy as a pattern-quality coordinate, require E.9.DA before ordinary E.21 first-pass reads, or reuse pattern-quality coordinates to judge the DRR as if it were an authored pattern body. | | C.25 | General Q-Bundle normal form. | Specialise Q-Bundle for FPF pattern quality. | Replace arbitrary engineering quality-family bundling. | | A.17, A.18, A.19, C.16 | Characteristic, scale, coordinate, measurement legality. | Use ordinal content readings and cite measurement law when live. | Convert ordinal readings into numeric measures, scores, percentages, or averages. | | F.18, E.10, A.6.P, C.2.P, C.16.P, C.16.Q | Naming, wording-use, relation, epistemic, characteristic/scale, and quality-term precision restoration. | Require recoverable kind, relation, characteristic/scale construction, quality-term endpoint, and admissible use for FPF-force-bearing heads. | Mint durable cross-pattern names by convenience, or let pattern-quality wording carry hidden score, scale, quality, evidence, gate, or release force. | | A.10, B.3, A.20, A.21, A.15 | Project evidence, assurance, local CV status, gates, work. | Keep project-side reuse under exact receiving patterns. | Certify project safety, compliance, gate passage, release, work authority, or publication truth. |

Coordinates with E.11 and J.4. E.21 is the governing pattern when the live question is: "Is this FPF pattern version good enough for this declared reader/use/scope, and may improvement stop?"

E.21 is not the first entry when the live question is only:

• how to write the pattern body (E.8);
• which admission/refresh review profile to run (E.19);
• how to frame the quality-read purpose before the pattern-quality read (E.22);
• how to repair entry discoverability or wrong-pattern selection (E.11);
• how to recover overloaded quality or characteristic/scale wording before a pattern-quality claim (C.16.Q, C.16.P, or C.25).

If E.21 changes Preface, J.4, ToC query rows, local Problem-frame recognition cues, or retrieval-facing entry support, the entry-facing effect remains governed by E.11; E.21 only supplies the pattern-quality claim.

Builds on:

• E.1 and E.2 for mission and pillar fit.
• E.8 for pattern authoring structure, recognition text, action guidance, and SoTA-Echoing obligations.
• E.19 for review and refresh profiles.
• C.25 for Q-Bundle normal form.
• C.16, A.17, A.18, and A.19 for characteristic, scale, coordinate, measurement, and characteristic-space discipline.
• F.18, E.10, A.6.P, C.2.P, C.16.P, and C.16.Q for naming, wording-use, relation, epistemic, characteristic/scale, and quality-term precision.

Coordinates with:

• C.2.P when epistemic precision repair touches FPF-force-bearing Problem frames, recognition text, examples, or worked slices; an E.21 quality read treats epistemic precision repair as incomplete when no remaining admissible reader move or exact neighbouring-pattern application survives.
• E.12, E.13, and E.14 for cognitive ergonomics, pragmatic utility, Goodhart/proxy checks, and human-centric working-model checks.
• C.18, C.19, and related NQD/OEE patterns when candidate fronts or search/archive constructs are live.
• G.11 when telemetry-driven refresh or decay is live.
• A.10, B.3, A.20, A.21, and A.15 when a pattern-quality result is reused as evidence, assurance, local CV status, gate-decision material, or work authority.

Constrains:

• Any FPF pattern-quality read that would otherwise use a single score, unscoped maturity label, or reviewer-taste verdict.
• Any pattern drafting or review stop decision that claims the current version is good enough for a declared use.

#E.21:End

Last Updated: 2026-05-29 — this section last modified in upstream FPF commit 2e112078 (github.com/ailev/FPF)