Findings Ledger

Stochastic judgment proposes; deterministic verification disposes (tier doctrine: review-tiers — .claude/rules/review-tiers.md). This skill is the proposes side's inbox: a single append-only JSONL at .claude/ledger/findings.jsonl where Tier 2 findings accumulate fingerprinted, so the same defect phrased two ways across runs usually collides to one entry (a heuristic — see the limits doc) and recurrence across independent runs — not rhetoric — decides what gets investigated.

The ledger file

• Location: .claude/ledger/findings.jsonl, one JSON event per line.
• Append-only. Status transitions are new events for the same fingerprint; prior lines are never mutated. Current status = the fingerprint's most recent event.
• Gitignored by default (machine-local noise). Team-shared mode: in .gitignore, replace the .claude/ledger/ line with .claude/ledger/*.lock (the lock sidecar stays local) and commit the findings file so recurrence is measured across the whole team's runs. Append-only + stable ordering keeps the diffs reviewable.
• CI runners don't persist it. Appends made during a CI job vanish with the runner — and CI must never push the ledger back to the repo. Upload the run's findings as a build artifact and harvest locally instead; patterns in references/ledger-format.md.

Entry fields: fingerprint, file, claim, tier, source, run_id, date, evidence (path or null), status (NEW | RECURRING | INVESTIGATING | PROMOTED | RETIRED-NOISE). Full schema, lifecycle, and fingerprint-normalization limits: references/ledger-format.md.

Commands

All via scripts/ledger.py (python3, stdlib only; exit 0 = ok, 2 = setup/usage error; output ordering is deterministic). Resolve the script path first — references/install-paths.md:

PROJ="${CLAUDE_PROJECT_DIR:-.}"
# 1. Repo checkout
LEDGER="$PROJ/.claude/skills/findings-ledger/scripts/ledger.py"
# 2. Project Codex install
[ -f "$LEDGER" ] || LEDGER="$PROJ/.agents/skills/findings-ledger/scripts/ledger.py"
# 3. Global Claude Code install
[ -f "$LEDGER" ] || LEDGER="$HOME/.claude/skills/findings-ledger/scripts/ledger.py"
# 4. Global Codex install
[ -f "$LEDGER" ] || LEDGER="$HOME/.agents/skills/findings-ledger/scripts/ledger.py"

python3 "$LEDGER" add \
  --file <path> --claim "<one sentence>" --tier 2 \
  --source <agent-name> --run-id <id> [--evidence <path>]

python3 "$LEDGER" tally
python3 "$LEDGER" triage [--threshold 2] [--age-days 14]
python3 "$LEDGER" promote <fingerprint> [--status INVESTIGATING] [--evidence <path>]
python3 "$LEDGER" retire <fingerprint>

• add computes the fingerprint (sha256 of file path + normalized claim text) and marks a repeat sighting RECURRING. A --tier 1 finding without --evidence is demoted to tier 2 automatically — the tier rule, enforced.
• Recurrence counts distinct run ids, not raw sightings — an agent repeating itself within one run cannot cross the threshold alone.
• triage lists fingerprints at/over the recurrence threshold as promotion candidates, and still-NEW single-run findings older than the age cutoff as retire-as-noise candidates. The /triage-findings command (repo-local maintainer tooling; not shipped by the installer) wraps this and proposes a ratchet target per candidate; the human disposes.
• promote refuses to record PROMOTED without --evidence — a promotion is its encoded check.

The ratchet

A recurring fingerprint is a candidate to leave the stochastic layer forever: investigate → encode as a scripts/validate.sh rule (Tier 0) or an exit-nonzero evidence script (Tier 1) → promote with the encoded check as --evidence → no LLM ever re-litigates it. Singles nobody re-reports get retired. The ratchet only turns toward lower variance.

Related skills

• code-review-and-quality — produces the findings this ledger stores
• skill-library-review — library-audit findings route here when unevidenced
• adversarial-claims-reviewer — the Tier 1 pattern: evidence artifacts instead of ledger entries