auditlisted

# Audit Run the full `make` audit in phases. Each phase builds on the previous; the ordering matters because the dev runner writes one archive per invocation, and granular runs must happen *before* their combined counterparts so you keep diffable per-tool logs. ## Repo context ```! cat .claude/skill-context.md 2>/dev/null || echo "(no .claude/skill-context.md — abort and direct the user to \`docs/conventions.md\` in the techne docs for the canonical scaffolding template)" ``` The injected content above is the source of truth for this repo's toolchain. Read the `## audit` section (and `## repo` for context) and use those phases — do not fall back to hardcoded defaults. Specifically, expect it to supply: - **Full-audit phase list** — ordered `make` targets, one per step, with a note on what each produces. - **Fast-audit subset** — which subset of phases is the "am I ready to push" probe. - **Stop-early phase** — which phase(s) aborting the rest if they fail. - **Log archive format** — glob and summary-block origin (e.g. `logs/dev-<ts>-<cmd>.log`, `SUMMARY` block). - **Do-not-run targets** — long-running / expensive targets the audit must skip. ## Modes - **Full audit (default)** — every phase from the injected list. Run when the user says "run the full audit" or didn't specify. - **Fast audit** — the fast subset from the injected list. Run when the user says "quick check", "fast audit", or only wants to know if lint + unit tests pass. Pick the mode up front and stick wi