Audit status

User Input

$ARGUMENTS

Interact with the dev in their working language — never force English on the user; the example phrases below are templates.

Context

Part of the PDD (Parity-Driven Development) method. This skill is read-only — it changes no files. Goal: know the overall project state without opening dozens of files.

Live alternative: the same data is available as a live CLI panel via pdd board (single snapshot) or pdd board --watch (re-renders when .audit/ changes). Mention this to the dev when it fits — /audit-status is the in-chat surface; pdd board --watch is the always-on terminal surface.

Outline

1. Initial checks

• Check whether .audit/BOOTSTRAP.md exists. If NOT: report "PDD is not initialized in this project yet. Run /audit-bootstrap first." and stop.
• Read .audit/BOOTSTRAP.md briefly to extract: Mission, Target date, PROJECT_AREAS, and CONFIDENCE_MIN (default tier-1 if absent).
• Read .audit/coverage.md if it exists (parity coverage map).
• Read .audit/board.md if it exists.
• List .audit/findings/ (all subdirs) — open findings.
• List .audit/resolved/ (all subdirs) — resolved findings.

2. Parse the findings

For each subdir in findings/ and resolved/:

• Read the YAML frontmatter of README.md. English keys: id, title, slug, area, severity, status, discovered-at, discovered-by, confidence, worktree.
  • confidence is one of tier-0 | tier-1 | tier-2 | tier-3.
  • worktree is an absolute path OR the literal none.
• Detect whether investigation.md exists (investigated flag).
• Detect whether resolution.md exists (resolved flag).
• List the contents of refs/ (evidence count).

3. Parse the coverage map

From .audit/coverage.md, parse the markdown table with columns:

| Behavior / Area | Reference case | Status | Tier | Finding |

• Status is one of not-started | finding-open | verified.
• Compute parity coverage = verified / total rows.
• If .audit/coverage.md does not exist, report coverage as "unavailable (run /audit-bootstrap to seed the coverage map)".

4. Apply filters (if $ARGUMENTS is present)

• No args: default overview.
• detailed: include an individual listing of each finding.
• area:<name>: filter by area (e.g. area:checkout).
• severity:<level>: filter by severity (e.g. severity:critical).
• open or investigated: filter by status.

5. Output — default format

═══════════════════════════════════════════════════════════════
  PDD Status · <project name>
═══════════════════════════════════════════════════════════════

Mission: <first line of the BOOTSTRAP Mission>
Target date: <BOOTSTRAP date> · <X days remaining or "no fixed date">
Confidence gate: <CONFIDENCE_MIN>

───────────────────────────────────────────────────────────────
  Parity coverage
───────────────────────────────────────────────────────────────

[███████████░░░░░░░░░]  <PP>%   (<verified> of <total> behaviors verified)
  └─ finding-open:  <F>
  └─ not-started:   <NS>

───────────────────────────────────────────────────────────────
  Summary
───────────────────────────────────────────────────────────────

Open findings:           <N>   (files in .audit/findings/)
  └─ already investigated: <M>
  └─ awaiting:             <N-M>

Resolved:                <R>   (last 7 days: <R7>)

───────────────────────────────────────────────────────────────
  Confidence distribution
───────────────────────────────────────────────────────────────

🟢 tier-3 (max · diff + characterization test):  <n>
🟠 tier-2 (high · automated data-to-data diff):  <n>
🟡 tier-1 (medium · paired screenshots):          <n>
🔴 tier-0 (low · textual description only):        <n>

<If any OPEN finding is below CONFIDENCE_MIN, add:>
⚠ <k> open finding(s) below the confidence gate (<CONFIDENCE_MIN>) — run /audit-compare to raise the tier.

───────────────────────────────────────────────────────────────
  By project area
───────────────────────────────────────────────────────────────

<For each area in PROJECT_AREAS from BOOTSTRAP:>
<area name>:  <n open> / <m resolved>

Other (uncategorized):  <n> / <m>

───────────────────────────────────────────────────────────────
  By severity
───────────────────────────────────────────────────────────────

🔴 critical:  <n>
🟠 high:      <n>
🟡 medium:    <n>
🟢 low:       <n>

───────────────────────────────────────────────────────────────
  In progress (from board.md)
───────────────────────────────────────────────────────────────

<contents of the "In progress" section of board.md>
<For each in-progress finding, if its worktree field is a path, show:  ↳ worktree: <path>>

───────────────────────────────────────────────────────────────
  Suggested next actions
───────────────────────────────────────────────────────────────

<conditional logic>:
- If there are open critical findings: "🚨 Prioritize NNN (critical) with /audit-investigate NNN"
- If there are investigated findings awaiting a fix: "There are M findings ready for /audit-resolve"
- If there are open findings below CONFIDENCE_MIN: "Raise evidence on NNN with /audit-compare NNN before resolving"
- If there are resolutions with an open PR: "Run /audit-qa NNN — PR is open and testable"
- If parity coverage < 100% and there are not-started behaviors: "Uncovered behavior: <first not-started row> — open it with /audit-new"
- If there are no open findings: "No open findings. All caught up."

Live view: pdd board --watch

6. Output with detailed

Append at the end:

───────────────────────────────────────────────────────────────
  Open findings (detailed)
───────────────────────────────────────────────────────────────

[NNN-slug]  <title>
  Area: X · Severity: Y · Discovered by @Z on YYYY-MM-DD
  Status: <open | investigated | out-of-scope>
  Confidence: <tier-N>  <🔴/🟡/🟠/🟢>
  Worktree: <path | none>
  Evidence in refs/: <N files>
  Path: .audit/findings/NNN-slug/

(repeat for each finding)

───────────────────────────────────────────────────────────────
  Resolved (last 7 days)
───────────────────────────────────────────────────────────────

[NNN-slug]  <title>
  Resolved on YYYY-MM-DD by @X · Confidence: <tier-N>
  Path: .audit/resolved/NNN-slug/

(repeat)

7. Rules

• Do NOT modify any file.
• Do NOT run build/test commands (this skill is purely reading from disk).
• If .audit/board.md is out of sync with the filesystem, report the inconsistency but do not fix it.
• If .audit/coverage.md is out of sync with the findings (e.g. a verified row whose finding is still open), report the inconsistency but do not fix it.
• Output must be scan-friendly — a dev running this skill is in "quick summary" mode.
• If PROJECT_AREAS is not filled in the BOOTSTRAP, group by each README.md's area frontmatter, or list as "uncategorized".

Quality rules

• NEVER invent numbers — if the info was not found, say "unavailable".
• ALWAYS read the BOOTSTRAP to get the current mission/target date and CONFIDENCE_MIN.
• If any finding has malformed YAML frontmatter, list it separately as "findings with a structural problem".
• The confidence distribution and parity coverage must be derived only from real data (findings' confidence field and coverage.md rows) — never estimated.