Whitaker ships an agent-facing skill under
skills/addressing-whitaker-findings/.
It distils the remediation patterns from adopting the suite across more than
forty repositories: what each lint expects, which fixes hold up under review,
when an exclusion is legitimate, and the traps that cost earlier adopters
rework. The skill is written for an agent rather than a human reader, and it
covers the lints described in Available Lints.
The skill lives in the repository rather than in a separate package, so it tracks the suite as the lints change.
A skill is a directory holding a SKILL.md whose YAML frontmatter is an Agent
Skills manifest. The manifest name is the discovery name a strict loader
uses, so a manifest that omits it leaves the skill undiscoverable. Every
shipped skill directory matches its manifest name, so one identifier names
the skill both on disk and at the point of discovery.
Making the skill available means copying its directory into the skills
directory that an agent tool reads. For Claude Code that directory is
~/.claude/skills/; other tools that consume Agent Skills use their own:
git clone --depth 1 https://github.com/leynos/whitaker.git /tmp/whitaker
mkdir -p ~/.claude/skills/
cp -R /tmp/whitaker/skills/addressing-whitaker-findings ~/.claude/skills/
The repository gates every shipped manifest, so a malformed one cannot reach a
copy taken from it: make lint runs skill-frontmatter-lint,
skill-manifest-validate, and skill-metadata-check over each manifest. The
Developer's Guide records how
those targets behave and when to run them alone.