audit-noise

Repository context. Gather first

Collect these with individual Bash calls, one command per call, never combined into a single invocation:

• Current branch, git branch --show-current
• Uncommitted .md files (empty = none matched or the probe returned nothing), git status --porcelain | grep -E '\.md"?$' | head -10

The pipe is the bound and belongs in the command. A read-time cap ("read only the first 10 entries") bounds nothing: the Bash tool returns the command's complete output into context before there is anything to decide about.

Treat a failure (not a repository, git unavailable) as an unknown value and carry on. Keep these as separate body Bash calls rather than pre-compute lines: the harness runs a skill's whole pre-compute block as one shell invocation, and a worktree-isolated session refuses a compound command that contains git. The dated record for that composition claim is the source-control plugin's worktree/reference/gather-block.md, "The pre-compute block runs as one shell invocation".

Pre-computed context

Noise findings (sample): !${CLAUDE_SKILL_DIR}/scripts/detect.sh 2>/dev/null | grep -E '^(status:|Summary total:|Finding shape:)' | head -20 || echo "none"

Purpose

Tracked markdown accumulates nine NOISE shapes distinct from FLAVOR, owned by the sibling /docs-hygiene:compress. The surfaces include rules, skill bodies, instruction files (CLAUDE.md, AGENTS.md), docs/, and READMEs. Each shape has to be kept current as the repository changes, and every reader pays to read past it in every file that carries it. This skill is a read-only classifier: it surfaces candidates with treatment guidance; the author hand-applies every edit.

Three of the nine, plan-reference, conversational-antecedent, and ticket-pr-residue, carry the same names the code-side sibling /code-tidying:audit-comment-residue uses, because they are the same authoring failure landing in a different file type. Ownership splits by file type, not by shape: markdown is this skill's, everything else is the sibling's, and neither scans the other's files. The patterns are not shared code. The sibling classifies only the extracted comment portion of a line; this skill classifies whole prose, where the same words carry meaning far more often, so its patterns are tightened accordingly and several of the sibling's cues are deliberately not carried over.