Finding a catalog skill
Every skill in the thrillmade/agent-skills catalog, grouped by what it
owns. Generated from docs/placement-map.json and gated: add, rename or
delete a skill without regenerating and validate-skills fails, so this list
cannot go quietly incomplete the way a hand-kept table does.
Read it before you write a rule. Before authoring a convention, a checklist, a house standard or a "how we do X" doc -- and before saying "nothing covers this". The catalog is bigger than the set installed in any one repo, so not seeing a skill is not evidence that none exists. Check here first; that check is one read, and writing the guidance again is not.
What a line is. The fragment after each name says what that skill owns, not what it says -- editorial text, gated for presence and length but not for freshness against that skill's own body. Enough to shortlist two or three; then open them. A skill you do not have yet:
npx skills add https://github.com/thrillmade/agent-skills --skill <name>
What this is not. Not a listing of your .claude/skills/ -- a repo holds
only what it subscribed to, plus anything it authored locally -- and not a
view of the wider skills.sh ecosystem. This indexes one catalog.
Reviewing a pull request
What a review flags, what it lets go, and what a claim has to cite.
api-contract-enforcement— public API shape changesbrand-voice-review— voice in user-facing stringscritical-issues-only— correctness, security, perf onlyevidence-based-review— quote the code or cut the claimguarding-a-regression— trusting a regression testpii-and-compliance— PII and secrets in logsproving-an-absence— claiming a thing is missingrespect-existing-conventions— a review is not a redesignskill-frontmatter-quality— judging a SKILL.md's frontmattertechnical-writing— docs and changelog for additionstest-discipline— test edits that hollow a suite
Running agents
Handing work out, gating what comes back, surviving a long session.
orchestrating-a-multi-agent-run— start here: the four obligationsorchestrating-agent-delegation— the brief, model tiering, verifyorchestrating-elite-agent-qa— the panel and design-critic gatesession-heartbeat— beats, checkpoints, resumingunattended-operation— what a handover permits
The catalog itself
Finding, censusing and authoring skills; decision logging; CLI conventions.
clud-bug-collaboration— living with the review botcurating-a-skill-catalog— the census rubric and verdictsfinding-a-catalog-skill— this directorylogmind— when and how to log a decisionretiring-a-superseded-decision— retiring a stale rulingtoken-frugal-tooling— quiet flags for the org's CLIs
Design: start here (L1)
Three entry points. Each routes to the primitives below in work order.
composing-a-screen— laying out a screenconsuming-a-design-system— using a system in a productdesigning-a-design-system— building or extending a systemreviewing-design-work— critiquing design output
Design primitives (L0)
Math and standards any design system can build on. No product opinions.
apca-contrast— Lc targets, the primary modelchroma-harmonization— per-stop cross-hue chroma capscomponent-sizing-principles— curated height and icon laddersdtcg-format— the W3C DTCG interchange fileempirical-design-principles— which design laws holdline-height-grid— two-track line height on gridmotion-continuity— motion between statesoklch-color-space— OKLCH ranges and gamut mappingpalette-relationships— hue-angle palette relationshipssemver-design-tokens— SemVer from the value diffspacing-system— minor+major units, 24px floortoken-naming-conventions— what belongs in a token nametype-scale— modular ratios and roundingusability-heuristics— what the heuristics saywcag-contrast— ratios, the legal cross-check
Web platform behaviour (L0)
What a shipping engine does that its spec and an emulator do not show.
ios-safari-surfaces— iPhone Safari bars, scroll, svh
Design review lenses (L0)
Judging the rendered surface -- and the standard it is judged against.
design-system-consistency— rendered drift from the tokensdesigning-elite-ui— the visual standard to build tofrontend-a11y— contrast, focus, targets, motionvisual-polish— the fine-but-not-elite lensweb-interface-guidelines-review— the code and markup lens
Design stances (L2)
One system's or one studio's lean on L0. The udts-* stubs are not guidance yet.
thrill-design— Thrill's house design tasteudts-component-sizing-ladders— per-density size laddersudts-dtcg-extensions— namespaced DTCG extensionsudts-linter-rules— machine-enforceable lint rulesudts-naming-convention— the concrete naming conventionudts-review— the system-specific lensudts-semver-defaults— SemVer policy choicesudts-spacing-defaults— density-mode spacing unitsudts-token-model— taxonomy and resolution
Deprecated (migration window)
Superseded. Each one names its successors; load those, not these.
component-sizing— use component-sizing-principlesdesign-token-naming— use token-naming-conventionsskillforge— use skill-creator / skill-smith
Templates
Copy-and-edit scaffolds. They fail the slug check until you customise.
conventions-template— a review-like-me scaffold
Deliberately not here
A map showing only what is inside teaches you the outside does not exist, which is the same mistake inverted. These are gaps this catalog knows it has. Each count is re-measured whenever this file is generated, over every skill's name and frontmatter description -- the surface that decides whether an agent finds it. A gap that had since been filled would refuse to render rather than go on advertising itself:
- TDD mechanics -- red before green, one failing test at a time. In the
superpowersplugin, not here --red before greenmatches 0 skills. - The same discipline under its other name --
test-drivenmatches 0 skills.
(Control: APCA matches 8, so the probe finds what is there.)
Not being named above is not evidence of absence either. Search skills/
before concluding a second time -- and control the search against a term you
know is there, the way the counts above are controlled.
Sources
- Published by thrillmot
- The catalog: https://github.com/thrillmade/agent-skills
- Placement, distribution and subscribers per skill:
docs/placement-map.json - The census that promotes, revises and retires these: https://github.com/thrillmade/agent-skills/tree/main/skills/curating-a-skill-catalog