audit-tests

What it does

Audit-tests is a read-only diagnostic skill that classifies the repository, maps applicable layers from a 7-layer testing taxonomy, runs deterministic quality gates through an in-repo audit-harness, builds RTM/personas/journeys traceability, and writes TEST_AUDIT.md plus observational sections of tests/TESTING.md. When P0/P1 gaps exist it mandatorily hands off to implement-tests instead of implementing tests itself.

How it works

• Enforces using the in-repo audit-harness for deterministic checks (hash-pin, escape-scan, CRAP, architecture, bias, Gherkin lint). Advises that the harness must travel with code and not reference user home paths.
• Step 1 reads tests/TESTING.md if present to drive classification and gap policy; verifies audit-harness installation in the target repo; runs manifest verification and branches logic for handoff decisions.
• Step 2 runs classify to emit audit-profile/v1 (union of repo classifications, gate set, and unresolved list), then uses test-discovery-agent to refine unresolved residues without overriding deterministic classification.
• Step 3 maps gates to layers using taxonomy-mapper-agent based on profile data, producing gap lists with P0/P1/P2 severity.
• Step 4 runs gates with commands like audit-harness conform, audit, scan, and currency, and then executes L3 tools (coverage, mutation, CRAP, architecture, bias, gherkin-lint). Blocking conditions are policy-driven; failures are recorded in TEST_AUDIT.md.
• Step 5 runs rtm-builder-agent, persona-coverage-agent, and journey-mapper-agent in parallel to produce RTM.md, PERSONAS.md, and JOURNEYS.md with coverage and gaps.
• Step 6 performs an escape-scan via escape-detection-agent on pending diffs, with exit codes 0/1/2 and REFUSE patterns recorded in TEST_AUDIT.md.
• Step 7 writes TEST_AUDIT.md (transient) with grade, classification, layer presence, gap lists, RTM/persona/journey summaries, and escape results; updates observational sections of tests/TESTING.md via Edit tool only.
• Step 8 if any P0/P1 gaps exist, builds a handoff payload and either prompts for confirmation on main/develop branches or proceeds autonomously on feature branches to trigger implement-tests.

When to use it

Use when auditing test quality, finding test gaps, or running the full 7-layer sweep. Trigger phrases include "audit tests", "find gaps", "full sweep", "7-layer audit", and related variants.

What it can touch

• Writes transient TEST_AUDIT.md at repo root.
• Updates observational sections of tests/TESTING.md, tests/RTM.md, tests/PERSONAS.md, and tests/JOURNEYS.md (through allowed observational edits only).
• Reads repository structure and existing TESTING.md; relies on in-repo harness installation and predefined scripts.

Caveats

• Relies on in-repo @intentsolutions/audit-harness for deterministic gates; performs no filesystem mutations beyond transient TEST_AUDIT.md and observational edits in testing docs.
• Hands off to implement-tests when P0/P1 gaps exist; main/develop branches require engineer confirmation before triggering implement-tests.
• Policies and thresholds are defined by the engineer-managed TESTING.md and related policy files; exit behavior follows those rules.