aoa-invariant-coverage-audit

name	aoa-invariant-coverage-audit
description	Audit whether existing tests and checks actually constrain the stable invariants that matter. Use when you need to judge if current coverage proves a rule or only repeats examples, and you want the smallest bounded follow-up gaps. Do not use when the invariant itself is still undefined or when the real task is to author new invariants rather than audit coverage.
license	Apache-2.0
compatibility	Designed for Codex or similar coding agents with repository file access and an interactive shell. Network access is optional and only needed when repository validation or referenced workflows require it.
metadata	{"aoa_scope":"core","aoa_status":"canonical","aoa_invocation_mode":"explicit-preferred","aoa_source_skill_path":"skills/core/engineering/aoa-invariant-coverage-audit/SKILL.md","aoa_source_repo":"8Dionysus/aoa-skills","aoa_technique_dependencies":"AOA-T-0017","aoa_portable_profile":"codex-facing-wave-3"}

aoa-invariant-coverage-audit

Intent

Use invariant-oriented coverage to judge whether an existing validation or proof surface really constrains the stable truth, and turn that judgment into a bounded audit package instead of a loose example review.

Trigger boundary

Use this skill when:

• an existing test, check, schema, fixture, generated/export parity check, report, receipt, or proof surface needs a review for invariant strength
• the question is whether current checks really constrain the stable rule
• you need to turn a loose example set into a bounded coverage audit
• you want an audit result that names the gap, not just the invariant
• a validation surface may look broad, but its claim limits and blind spots need review

Do not use this skill when:

• the main problem is defining the invariant itself rather than auditing coverage; use aoa-property-invariants first
• the invariant itself is still unknown and you need discovery work first
• the task is mostly about presentation details or a narrow snapshot
• you need a full boundary contract review rather than a coverage audit

Inputs

• target rule or stable truth
• current tests, checks, schemas, fixtures, reports, generated parity checks, proof reports, or examples
• known edge cases and stress cases
• the input space or states that matter most
• known claim limits or downstream surfaces that rely on the invariant

Outputs

• invariant coverage map
• gap list for weak or missing checks
• bounded follow-up checks or revisions
• concise verification summary
• audit verdict on whether coverage is strong enough for the current stable truth
• claim-limit notes that say what the current validation surface does not prove

Procedure

1. name the stable truth in plain language
2. map each existing check to the invariant it constrains
3. when the validation surface is not an ordinary test suite, choose the smallest useful shape from references/coverage-audit-shapes.md
4. mark weak, redundant, or missing coverage
5. add only bounded cases that strengthen the weakest invariant first
6. separate the audit result from any invariant-authoring work that should happen elsewhere
7. report what the suite or validation surface now proves and what it still does not

Contracts

• every claimed invariant must be traceable to at least one check
• the audit should distinguish real constraint from mere example repetition
• schema, report, generated, fixture, and proof surfaces should be audited by the stable claim they constrain, not by how complete they look
• the result should stay reviewable without sprawling into a full test strategy
• the skill should stay an audit package rather than a general testing doctrine
• the output should make the next coverage move obvious
• claim limits should remain visible when downstream readers may over-trust the surface

Risks and anti-patterns

• counting examples as if they were invariants
• letting random data hide thin coverage
• treating schema validity, report completeness, or generated freshness as proof of the underlying invariant
• widening into generic test design or full suite architecture
• making the audit so abstract that no one can act on it
• drifting into invariant authoring when the task is really coverage review
• turning the audit into a generic quality sermon instead of a bounded package

Verification

• confirm each core invariant is named and mapped to a check
• confirm the coverage gap list is specific and bounded
• confirm non-test validation surfaces are judged by their constrained claim and claim limit
• confirm another human can follow why the existing surface is or is not sufficient
• confirm the result names the next bounded follow-up, if one is needed
• confirm the skill reads like an operational audit rather than a plain technique lift

Technique traceability

Manifest-backed techniques:

• AOA-T-0017 from 8Dionysus/aoa-techniques at fbead87e01b82df6c56e3d92a074cd7515131847 using path techniques/proof/skill-support/property-invariants/TECHNIQUE.md and sections: Intent, When to use, When not to use, Inputs, Outputs, Core procedure, Contracts, Risks, Validation

Adaptation points

Project overlays should add:

• local test or check commands
• domain-specific invariants and edge cases
• local schemas, generated parity checks, proof reports, receipts, or fixture families that should be audited through references/coverage-audit-shapes.md
• notes on where coverage gaps are recorded
• rules for when a coverage audit should escalate into a deeper review