Menu
Auto-fix markdown lint issues across all .md files and report what changed
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Verify the project's hard memory + storage budgets — definite-leak-free under valgrind, peak host-heap below the v2 ceiling, ESP32 flash + RAM headroom non-negative, and (informational) RTC-Slow-Memory budget. Use when the user asks to "check budgets", "verify no leaks", "measure peak heap", "is the firmware too big", or as the §9 Heap-Profiling gate before tagging v2.0.0. Do NOT invoke as a side-effect of an unrelated build/test failure — it costs ~5–10 minutes wall-clock.
Convert PGM dumps (from `make test-native`) to PNG using the existing `scripts/pgm-to-png.py` converter — never reach for `convert`/`magick`/`ffmpeg`/PIL one-liners/`for`-loops. The script already handles batch mode (whole `.tmp/v2-pgm/`), single-file mode, and `--bands` for pixel-band Y-ranges. Use whenever the user asks to convert, visualize, render, or "look at" PGM output, or asks for pixel-band/Y-range info.
Check whether a CLI tool or package is available; if not, ask the user to install it before continuing
Run clang-format (auto-fix) across every C/C++ file in the repo and report what was changed; then run cppcheck and surface any remaining static-analysis warnings as a punch list. Mirrors fix-markdown — formatter auto-fixes, linter reports. Use when the user asks to "fix the C++ lint issues", "format the C++ code", or "run the C++ linters". Do NOT invoke as a side-effect to clear an unrelated hook failure.
Cut a new tagged release of bustaferl. Enforces the Tier 2/3 test-gate (host + device + soak), composes release notes from git log, and creates an annotated tag. Use when the user wants to "tag a release", "ship v1.x", "cut a release", or asks to verify pre-tag readiness. Optional `--major` flag adds the 24h test gate. Refuses if tests are stale or rot, blocks force-push to main.
| name | fix-markdown |
| description | Auto-fix markdown lint issues across all .md files and report what changed |
Run markdownlint-cli2 --fix across every Markdown file in the repo
(matching the pre-commit hook's glob) and report what was changed.
This skill is deliberately project-wide. It edits every .md file
that has a fixable lint issue, not only files you intend to commit. In
a repo where multiple Claude Code sessions can run concurrently, that
sweep can clobber another session's working-tree edits if the fixer
touches a file they have open.
Only invoke this skill when:
.md file.Do not invoke it as a side-effect of another flow to clear an
unrelated hook failure — that is the anti-pattern the /commit
skill's three-check protocol exists to prevent. Touching files you
have no intent to commit violates the project's "commit only your own
work" rule.
Record which .md files currently have unstaged changes — these
are the before set, used in step 3 to detect what the fixer
touched:
git diff --name-only -- '*.md'
Run markdownlint-cli2 in fix mode against the same glob the
pre-commit hook uses, so the skill and the hook agree on scope:
markdownlint-cli2 --fix '**/*.md' '!node_modules' '!.venv' '!.claude/security-review-latest.md'
(There is no make lint-markdown target in this repo — earlier
versions of this skill referenced one that did not exist.)
Re-run the git diff --name-only -- '*.md' from step 1 to get the
after set. Files newly present in the after-set were modified by
the fixer. Files already present in both sets may have had
additional fixer-applied changes layered on top of pre-existing
unstaged edits — disentangling that requires per-hunk inspection
and is not part of this skill's job.
Report:
If files were modified, remind the user to review the fixer's
changes before staging them. Stage and commit via /commit per
the project's commit protocol — never git add directly.
Do not stage or commit — leave that to the user.