An adversarial review skill that forces Claude into three hostile reviewer personas so 'LGTM' is never an option.
Reviews a diff or file purely for what can be deleted or simplified — over-engineering, speculative abstractions, and reinvented stdlib.
Forces a collaborative requirements interview, 2–3 alternative designs, and a committed spec document before any code is written.
Converts a raw git log into grouped, user-facing release notes with a headline and migration note.
Detects your project's stack and auto-configures lint, typecheck, test and format quality gates plus a Claude settings file.
Finds every CLAUDE.md in your repo, grades it against a quality rubric, and applies only the targeted improvements you approve.
A read-only skill that scans a diff, file, or repo for both over-engineered code and bloated prose, returning a ranked list of what to cut.
A mode skill that strips filler from Claude's prose and enforces YAGNI-first, minimum-code decisions.
Generates Conventional Commits messages with ≤50-character subjects that explain the "why" instead of restating the diff.
Turns a conversational bug report into a durable GitHub issue written in the project's own domain language.
Turns raw git commit history into categorized, customer-friendly release notes in minutes.
Scans any repository and produces an annotated ASCII directory tree plus a module table and suggested entry points.
A persona skill that forces the laziest solution that actually works — stdlib and native features first, shortest diff, no unrequested abstractions.
A skill that scaffolds a new Angular application with the Angular CLI and enforces modern Angular code-generation practices afterwards.
Forces every code review comment into one line — location, problem, fix — ready to paste into a pull request.
Turns your Git history into a categorized CHANGELOG.md with a Semantic Versioning bump suggestion.
A one-shot quick-reference card listing every Chisle command, rule and safety boundary.
An 8-phase, approval-gated pair programming workflow: understand requirements, write a spec, implement in vertical slices, review, then keep spec and code in sync.
Defines how the agent should interpret and act on blocking warnings from the comment-checker hook that runs after file edits.
A read-only skill that reports file metadata and statistics — size, modification time, line counts, function/class counts — without touching your files.