An audit pass for SKILL.md files that strips rationale prose, hardens soft language into imperatives, adds STOP: gates, and removes redundancy.
Creates new Lattice atoms, molecules, and refiners with the correct files, sections, and conventions.
A lint skill that structurally validates Claude Code agent and skill markdown files — frontmatter, naming, and body rules — so misconfigured files don't fail silently at runtime.
Generates a structured album progress dashboard with per-phase completion percentages, a track status table, and blocking items.
Enforces the edit → eval → lint → test → commit workflow when modifying skills, agents, hooks, or the eval framework inside the claude-elixir-phoenix repo.
A step-by-step release runbook that syncs the version across every manifest, builds, tags, publishes a GitHub release, regenerates the changelog, and notifies Discord.
Generates and audits accessible alt text for figures in R package (pkgdown) sites and Quarto documents by reading the plotting code and surrounding prose.
A persona skill that teaches Kadam/lamrim foundations — the three scopes, bodhicitta, reliance on a teacher — in Atiśa's voice, with mandatory Tibetan-canon citations.
Answers Theravāda questions — Visuddhimagga, Abhidhamma, Pāli suttas — in Buddhaghosa's commentarial voice, with mandatory source citations.
A hands-on reference for building Claude Code slash commands — frontmatter, dynamic arguments, file references, and plugin patterns.
A meta-skill that walks you through designing, writing, and validating Claude Code skills with a 6-step process and checklists.
Reference skill for uploading media from URLs or local paths to BulkPublish and batch-creating posts across social platforms.
A reference skill that teaches Claude Code plugin directory layout, plugin.json manifest fields, and component auto-discovery rules.
A reference skill for auditing web content against WCAG 2.2 and prioritizing accessibility fixes.
Reviews changed Java code straight from your git worktree and writes a severity-graded review.md with concrete fixes.
For every reviewer question thread on a PR, it recovers the original implementation reasoning and drafts a plain-text answer.
A prioritized ruleset that makes Claude write minimal new code that blends seamlessly into your existing codebase.
Measures your note vault's link graph and reports orphan rate, average degree, components, hubs, bridges and clusters — plus what each number means for retrieval.
Turns Claude into an orchestrator on explicit /director invocation: it defines, divides, decides and reviews while lower-tier agents do the actual work.