<< All versions

Skill v1.0.0

currentAutomated scan96/100
nomadicdaddy/aidd/execute-audit
──Details
PublishedSeptember 27, 2026 at 11:18 PM
Content Hashsha256:bcfdcece4720970e...
Git SHA
──Files
Files (1 file, 4.3 KB)
SKILL.md4.3 KBactive
SKILL.md · 84 lines · 4.3 KB

version: "1.0.0" name: execute-audit description: 'Run current aidd audits against a local project, preserving reports under .aidd/audit-reports/ and actionable findings as .aidd/features/audit-*. Use to run named audits or the profile-applicable audit set and capture their findings.' metadata: aidd-category: audit-remediation aidd-references: audits/SEVERITY_CLASSIFICATION.md


Execute Audit

Run one or more aidd audits against a local project through audit mode, preserving reports under .aidd/audit-reports/ and actionable findings as .aidd/features/audit-*.

Usage

execute-audit <app> <audit-name>[,<audit-name>...]
execute-audit <app> --audit-all

Arguments

  • <app> - Existing local project directory. A bare name resolves below aidd's configured

applicationsRoot; a relative or absolute path resolves directly.

  • <audit-name> - One audit definition name from <aidd-root>/audits/, without .md, or a

comma-separated list of names.

  • --audit-all - Run missing or stale audits from the profile-applicable set. aidd ranks that set

by change potential; it does not run in filename order.

Workflow

  1. Resolve the app to an existing local directory. Do not require a pre-existing .aidd/: audit

execution creates the metadata directories and stages the selected audit definitions.

  1. Verify every explicitly requested audit exists and is runnable under <aidd-root>/audits/.

For --audit-all, let aidd discover runnable definitions, apply the project profile and overrides, skip fresh reports, and rank the remaining audits.

  1. Run from <aidd-root>:

``bash bun run start -- --project-dir <app-dir> --audit <AUDIT_NAME>[,<AUDIT_NAME>...] ``

For the profile-selected set, run:

``bash bun run start -- --project-dir <app-dir> --audit-all ``

  1. Let aidd parse AIDD_RESULT and persist accepted reports, findings, finding-ledger events, and

any resulting roadmap assignments. Do not hand-write .aidd/audit-reports/, .aidd/features/, .aidd/findings-ledger.jsonl, .aidd/roadmap.json, or .aidd/CHANGELOG.md outside that contract.

  1. After the run, inspect the summary and generated paths. Expected outputs are:
  • .aidd/audit-reports/<AUDIT_NAME>-YYYY-MM-DD.md for every accepted report (a second run of

the same audit on the same day replaces that audit's report);

  • zero or more .aidd/features/audit-* feature directories;
  • .aidd/findings-ledger.jsonl events when the run reports findings;
  • an updated .aidd/roadmap.json when new findings are assigned into an existing roadmap.
  1. If new feature files were created, read them back to confirm their IDs, content, and count.

aidd validates the complete feature collection after each iteration and reports contract issues in the run; do not shell into the aidd installation to run --check-features again.

  1. Review prior reports before launching. Audit mode automatically supplies up to three recent

.aidd/audit-reports/<AUDIT_NAME>-*.md summaries to the auditor; also check docs/audits/<AUDIT_NAME>_AUDIT_REPORT_*.md read-only as a legacy fallback (do not write new reports there). This matters because a same-day run replaces the current dated report.

  1. Classify repeated findings from aidd's persisted result:
  • an open fingerprint match is a pre-existing finding and is suppressed as a duplicate, not a

regression;

  • a fingerprint previously remediated or otherwise resolved and then emitted again is a

recurrence, recorded in .aidd/findings-ledger.jsonl and in the new feature's notes;

  • a finding previously dismissed as a false positive remains suppressed.
  1. Report the audit names, report paths, total and newly created finding counts, validation result,

suppressed duplicate/dismissal counts, recurrences, and remaining audits. For a confirmed recurrence, recommend an appropriate automated prevention mechanism such as a lint rule, CI assertion, smoke:qc check, or targeted regression test.

Notes

  • Severity definitions live in .aidd/audits/SEVERITY_CLASSIFICATION.md (staged; or <aidd-root>/audits/SEVERITY_CLASSIFICATION.md in the aidd repo).
  • Use bun run aidd-tools -- audit:profile-mapping when audit applicability mapping appears stale.
  • Audit reports live exclusively at .aidd/audit-reports/.
All versions