Appearance
Skills
The 53 skills we author, grouped by what you reach for them to do — the directory name is the command you type.
Plan & decide
Shape and pressure-test an idea before any code — widen the options, judge whether it's worth doing, and red-team the thinking.
brainstorm
standalonewriteshands-on
Use when a fuzzy idea needs shaping into a spec before any code
Read morecheck-reasoning
checkreadhands-off
Use when a plan or diff needs red-teaming before you commit
Read moregrill-with-docs
grillwriteshands-on
Use when you want a plan tested against your domain docs + the docs updated
Read morevet
standalonereadhands-off
Use when you're unsure an idea or tool is worth adopting — and where it belongs
Read morewayfinder
standalone
(from mattpocock/skills v1.1, kept close to upstream as an experiment) Plan a chunk of work too big for one agent session as a shared map on GitHub — a `wayfinder:map` issue with child investigation tickets, native blocking edges, and a fog-of-war "Not yet specified" section — then resolve ONE ticket per session until the way to the destination is clear. Use to try the wayfinding flow on a big, fuzzy, multi-session effort. NOT brainstorm (single-flow double diamond → spec) · NOT organise (places the existing backlog) — this charts decisions still in fog.
Read moreBacklog
Turn plans into typed issues, classify them for readiness, place them into epics/sprints/milestones, and see what's next.
find-untracked-work
findwriteshands-off
Use when you want built-or-intended work that isn't tracked surfaced
Read moreorganise
standalonewriteshands-on
Use when issues need placing and ordering into epics / sprints / milestones
Read moreto-blog-issue
to
Capture a blog-post idea as ONE GitHub issue labelled `Blog post` — a braindump an agent later drafts from in the house voice. Files one backlog issue: the angle, why it matters, points to cover, source material. Surface-then-confirm before filing. Use for "this could make a good blog post", "blog this", "capture a blog idea". NOT to-tickets (a typed tree of build slices); NOT a drafting step (the post is written later, from the issue).
Read moreBuild
Execute the work — scaffold features and skills, and run sprints and epics one gated commit at a time.
build-helper-step-in-tempo
buildwriteseither
Use when you're adding a Task Helper enricher or refiner
Read moreShip & review
Get changes out the door — plan and action the open-PR stack, and roll one review nit out everywhere it occurs.
build-from-qa-plan
buildwriteshands-on
Use when you want the doable items off a QA plan actioned
Read morebuild-pr-feedback
buildwriteshands-on
Use when you want one review nit applied everywhere it occurs
Read morebuild-qa-plan
buildwriteshands-off
Use when you want the open PRs for a sprint or epic planned
Read moreVerify & QA
Prove a change is right before it ships — find missing tests, check conformance, root-cause bugs, and hunt vulnerabilities.
check-everything
checkwriteshands-on
Use when you want a full health sweep of your current change
Read morepreflight
standalone
The make-green loop — run ci:check, fix, re-run, then chase mutation survivors — is the noisiest churn in a session and it floods your context. This runs the whole loop in an isolated fork and hands back one verdict, fixing the mechanical failures itself and flagging only the real decisions, so the thrash never reaches your thread.
Read moreOptimise
Make an existing thing faster or leaner without changing what it does — code, CI, prompts, the always-on context.
Harness care
Keep the harness itself healthy — audit the skill set and the whole system, measure whether a skill does what it promises, reconcile the docs, capture learnings, and source improvements.
eval
standalone
Changing a skill or mode used to be a vibes call — a tweak that looked better might silently regress a rule, like caveman dropping issue-links under compression. This runs a harness eval for one skill: control-vs-treatment sub-agents over its co-located eval set, blind-graded against a weighted rubric, reporting the delta and a hard PASS/FAIL on the leak-canary and link guards — so a change is measured, not guessed.
Read morefind-harness-improvements
findreadhands-off
Use when you want ideas to improve the whole harness
Read moreWays of working
Standing manners that change how the agent works for a whole stretch, not one-off actions.
Housekeeping & reference
Tidy-ups and reference — repo cleanup, end-of-session wrap-up, session transcripts, the domain conventions, and frontend design taste.
build-transcript
buildreadhands-off
Use when you want this session exported for analysis or learning
Read moredesign-taste-frontend
designwriteseither
Use when you're building a frontend that mustn't look templated
Read more