Site Claim Guardrails

Public claim levels

Claim levels govern wording strength, not product readiness.

shippedMain-true source behavior or source-document guidance has a public-safe proof path, rule ID or family, coverage label, limitation, and source context.

demoChecked-in public demo proof, sanctioned demo artifacts, or public-safe demo summaries support the exact wording and keep limitations visible.

conceptFuture-facing, explanatory, dev-only, process-level, or not-yet-backed wording stays framed as guidance and cannot be presented as shipped or demo-backed.

hiddenPrivate, unreleased, unsafe, or local-only detail is omitted or abstracted until an owner provides a public-safe summary.

Page-level boundaryThe concept label on this page does not upgrade any row into shipped behavior, demo-backed behavior, impact proof, runtime proof, release-safety proof, or complete coverage.

Proof-path requirements

A public claim needs all supporting fields before it grows stronger.

proof pathPoint to a public-safe route or artifact summary, with source context such as main, public demo, future-only, dev-only, hidden, or local-only.

rule basisUse a public-safe rule ID when available; otherwise use a rule family and a limitation explaining why the exact rule is unavailable.

evidence tierUse only Tier1Semantic, Tier2Structural, Tier3SyntaxOrTextual, or Tier4Unknown when a tier applies.

coverage labelTranscribe reduced, partial, unknown, unavailable, future-only, local-only, or hidden labels without normalizing them into stronger wording.

limitationNo claim is upgraded by confidence, seniority, repetition, urgency, roadmap intent, pressure, or appealing phrasing.

Allowed evidence references

Reference public-safe summaries and route families, not private proof.

Use public surfacesAllowed references include public pages, documentation, rule catalog pages, public-safe generated summaries, sanctioned demo artifacts, review packets, route metadata, coverage labels, limitations, and safe extractor-version labels.

Name artifact families only as categoriesFact streams, SQLite indexes, analyzer logs, reports, and rule catalog entries may be named as boundary categories, not linked as raw public proof.

Preserve the evidence boundaryKeep rule ID or family, evidence tier when applicable, coverage label, limitation, and source context beside the public statement.

Private-only supportPrivate-only evidence can support internal follow-up, but public copy needs a public-safe summary before it cites that support.

Forbidden raw material

Raw or private material stays out of public copy.

Do not publish raw proofForbidden categories include raw facts, raw SQLite databases, analyzer logs, source snippets, raw SQL, config values, secrets, local paths, remotes, generated scan directories, private sample names, command output, hidden validation details, and credential-like values.

Apply this everywhereThe boundary applies to rendered text, decoded HTML attributes, metadata, sitemap or discovery output, examples, fixtures, validation messages, and review-packet references.

Use abstract labelsHidden work is one abstract row or omitted; it does not reveal private names, internal route names, counts, cadence, sequencing, or private proof details.

Non-claim patterns

Say what static evidence does not establish.

Static boundaryTraceMap static evidence can show a bounded relation, gap, coverage label, or owner handoff without proving impact, absence of impact, safety, approval, business correctness, or production behavior.

Do not sayTraceMap proves runtime behavior, production traffic, endpoint performance, outage cause, release approval, release safety, operational safety, complete coverage, AI/LLM impact analysis, or replacement of human review.

Prefer neutral wordingUse phrases such as evidence shows, coverage is reduced, not established by this scan, needs review, owner follow-up needed, internal only, downgrade, hidden, and stop condition.

Downgrade and hidden rules

Choose the row before repeating the sentence.

Public wording guardrail rows. Each row includes the required condition, allowed action, proof path, downgrade trigger, forbidden implication, and review handoff.
Row	Condition	Allowed public wording/action	Required proof path	Downgrade/hidden trigger	Forbidden implication	Review handoff
shipped	Main-true behavior or source-document guidance exists.	Repeat only with proof and limitations.	Public-safe source, rule or rule family, coverage label, and limitation.	Missing proof field or reduced evidence.	Runtime, release, safety, complete coverage, or replacement of review.	owner follow-up needed
demo	Checked-in public demo proof or public-safe demo summary exists.	Repeat as demo with proof.	Public demo proof, evidence boundary, coverage label, and limitation.	Demo proof is absent, private, raw, or stale.	Shipped, production, runtime, or release wording.	downgrade before repeating
concept	Future-facing, guidance, dev-only, or not-yet-backed wording.	Keep concept wording.	Public-safe concept source and limitation.	Wording implies shipped or demo-backed behavior.	Public availability, operational readiness, or release timing.	downgrade before repeating
hidden	Details are private, unreleased, or unsafe to disclose.	Omit or abstract.	None public; use hidden rationale internally.	Any public detail would expose private material.	Hidden names, counts, cadence, sequencing, or in-flight status.	hidden
raw artifact reference	A claim cites raw or private material.	Replace with a public-safe summary.	Public-safe summary or owner-approved public artifact.	Only raw/private proof exists.	Direct raw facts, SQLite, logs, snippets, SQL, config, paths, remotes, command output, or hidden validation details.	owner follow-up needed
dev-only feature	Evidence exists only off public/main surfaces.	Treat as concept or hidden.	Public-safe dev/context note if publishable.	Wording implies availability.	Shipped, demo, timing, or production use.	internal only
reduced coverage	Evidence is partial, reduced, unknown, unavailable, or gap-labeled.	Keep reduced coverage visible.	Public-safe artifact with coverage label and limitation.	Claim depends on missing coverage.	Clean, complete, safe, or no-impact wording.	owner follow-up needed
runtime/release wording	Copy implies runtime behavior or release approval.	Remove or rewrite.	Not supported by TraceMap static evidence.	Any runtime or release assertion appears.	Production traffic, endpoint performance, outage cause, release safety, or operational safety.	do not repeat
AI/LLM wording	Copy says core scanner/reducer uses AI/LLM analysis.	Remove or rewrite.	Deterministic static evidence only.	AI/LLM impact-analysis wording appears.	Embeddings, vector search, prompt classification, or AI impact analysis.	do not repeat
private-only support	Support exists only in private or local evidence.	Internal only, hidden, or owner follow-up.	Public-safe summary required before publication.	No public-safe summary exists.	Public proof, capability, availability, or customer-specific implication.	internal only

Validation expectations

Validation checks the route, not hidden proof.

Route and metadataThe standalone route needs concept-level title, description, canonical URL, Open Graph metadata, sitemap metadata, and discovery metadata.

StructureChecks cover visible concept level, the shared principle, required anchors, adjacent links, row markers, row fields, and body word-count bounds.

Public-safety sweepValidation scans rendered text, decoded attributes, raw HTML, metadata, sitemap, discovery output, tests, examples, and messages for forbidden claims or private/raw material.

Browser sanityDesktop and mobile sanity checks confirm the public page is readable and the table remains reachable without adding the route to primary navigation.

Review handoff

Use bounded next states when proof is missing.

repeat with proofThe public-safe proof path, rule basis, tier if applicable, coverage label, limitation, and source context are present.

downgrade before repeatingThe wording is stronger than the proof and must move to concept, demo, reduced, or narrower language.

owner follow-up neededAn owner must provide a missing proof path, public-safe summary, limitation, source context, or decision.

do not repeatThe sentence makes a forbidden implication or lacks the minimum proof fields.

internal onlyPrivate-only or raw-only evidence can guide internal follow-up but is not public proof.

hiddenPublishing details would expose private material, hidden work, or unsafe context.

Adjacent surfaces

This page sets copy guardrails; nearby routes keep their own jobs.

Review claim checklistCanonical ritual for deciding whether a specific sentence can be repeated, downgraded, sent to an owner, withheld, or kept internal.

Proof source catalogRoute-to-source mapping for public-safe proof sources; this page only says what categories may be referenced.

RoadmapClaim gates and maturity context without adding timing promises here.

LimitationsBroader product and analysis boundaries; this page shows when those boundaries force downgrade, hiding, or owner follow-up.

Questions and objectionsStakeholder Q&A; this page teaches authoring boundaries.

Change-risk language guideWording patterns for static change-risk language; this page defines claim-level and publication rules.

Keep public TraceMap copy inside its proof boundary.