plan-work
49Skill Claude Code
Decompose a chunk of the approved plan into a forward-only DAG of phases and tasks, write it to plan.edn, return a compact summary.
Skill Claude Code
Decompose a chunk of the approved plan into a forward-only DAG of phases and tasks, write it to plan.edn, return a compact summary.
Skill Claude Code
Write an ADR when an architecture decision is made or work settles a real choice between alternatives.
Skill Claude Code
Run a blameless incident review with action items.
Skill Claude Code
Surface risky assumptions and missing acceptance criteria.
Skill Claude Code
The authoritative procedure for one review round: lanes, reviewer fan-out, spine triage, level-ordered editor waves, integrate, verify, advance.
Skill Claude Code
Run a risk spike: name the load-bearing unknown, time-box it, build a throwaway harness that exercises the seam at and beyond expected load, decide go or no-go or mitigate, feed the outcome forward.
Skill Claude Code
Cut a release end to end, from lane gate to tag and trunk advance.
Skill Claude Code
Turn noisy signals into structured analysis.
Skill Claude Code
The verification lane table. Runs the descriptor's three lane tiers in order and reports pass or fail per lane. Invoked by the verifier agent and the review-round-runner.
Skill Claude Code
After a major slice lands, root-cause its review findings and its token and verification cost, then commit improvements back into the skill system so the next slice is cheaper and better.
Skill Claude Code
Recipe for writing C in the runtime, module placement and the ownership and control-flow discipline. Invoked by writer agents when a C unit is dispatched.
Skill Claude Code
Grow a differential conformance corpus with per-target edge forms, free palette tier first, judgment tier for the reasoning-gated remainder. Invoked by writer agents when a conformance generation unit is dispatched.
Skill Claude Code
Classify differential-probe divergences into real bug, intentional divergence, or harness artifact, and route each to its owner. Invoked by reviewer agents over one module shard of diff output.
Skill Claude Code
Recipe for changelog entries, when a change deserves one, the entry voice, the category-first form, and the release-cutting ritual. Invoked by writers and editors when authoring entry lines, and at release time.
Skill Claude Code
Recipe for writing Clojure, Functional Core / Imperative Shell at the function level (namespaces by domain), native wrappers, and the boundary discipline. Invoked when writing Clojure for the project.
Skill Claude Code
Recipe for any commit. Mandates jj over git, the single-line category-first message, and the jj workflow. Invoke before any commit, by any agent or the maintainer.
Skill Claude Code
Recipe for writing Elixir, pure functions in modules as the core and GenServer plus OTP as the shell, supervision trees, and NIF discipline for the C and Zig edge. Invoked when writing Elixir for the project.
Skill Claude Code
Recipe for any English the agent produces in this project, commits, comments, docstrings, ADRs, design docs, skill bodies, changelog, error messages, project guides. Optimizes for terse, humanized prose with no em dashes and no AI tells. Invoke for any prose-writing activity.
Skill Claude Code
Recipe for writing tests, surface selection across pure core, native edge, shell, and full vertical, property and generative discipline, and the teeth rules. Invoked by writer agents when a test unit is dispatched.
Skill Claude Code
Recipe for writing UI as data, the pure layout and hit-test functions, the widget vocabulary, the renderer diff contract, and the design language as a working discipline. Invoked when writing or changing any UI surface.
Skill Claude Code
Recipe for writing Zig, native bodies and edge wrappers, the allocator and lifetime discipline, and the hot-path rules. Invoked when writing or editing any Zig source.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: