Univers42
Organizationconfiguration of claude code here you have the best practice of skillls, hooks, rules, cripts, and different tools to help the ai consume much less tokens.
Categories
Indexed Skills (13)
brainstorm
Generate options properly — diverge wide before judging, then converge on evidence and cost. Ends with one recommendation and a kill criterion, not a list. Auto-triggers on: "brainstorm", "what are the options", "how else could we", "give me ideas", "explore approaches", "what would you suggest"
browser-testing
Verify a change in a real browser and come back with evidence — navigate, interact, snapshot, read the console. Turns "should work" into an artifact. Auto-triggers on: "test in the browser", "does the page work", "check the console", "screenshot", "click through", "playwright", "verify the UI", "e2e test"
commit-craft
Turn a working tree into a history someone can read, revert and bisect — atomic commits, Conventional Commits, never co-authored. Auto-triggers on: "commit this", "write a commit message", "split these changes", "prepare a PR", "clean up the history"
context-budget
Measure and cut what this config costs in context every session — always-on rules, skill descriptions, unused skills. The token-reduction pass. Auto-triggers on: "reduce token usage", "context is full", "what is loading", "trim the config", "context budget", "why is my context so big", "compacting too often"
debug
Find the actual cause of a failure instead of guessing at fixes. Reproduce, bisect to the smallest failing case, prove the mechanism, then fix once. Auto-triggers on: "why is this failing", "debug this", "this test is flaky", "it works locally", "fix this bug", "this crashes"
design-review
Judge an interface the way a design engineer does — hierarchy, rhythm, type, states, and the empty/loading/error cases nobody built. Names what is wrong and why, not "make it pop". Auto-triggers on: "review this design", "does this look right", "improve the UI", "critique this", "the design feels off", "make this look better"
doc-sync
Find the documentation a change just made false, and fix it — examples copied from passing tests, never composed. Auto-triggers on: "update the docs", "are the docs current", "the README is out of date", "document this change", "sync the documentation"
frontend
Build UI that holds up — component and state boundaries, design tokens, responsive and theme behaviour, and the accessibility gate. Auto-triggers on: "build a component", "add a page", "style this", "make it responsive", "dark mode", "the layout breaks", "frontend", "this UI"
originality
Prior-art pass before writing something new — find what already does this, in the repo and outside it, then say plainly whether to reuse, wrap, or build and why. Auto-triggers on: "has this been done", "is there a library for", "should I build this", "prior art", "am I reinventing", "write a X from scratch"
perf-budget
Set the number before you optimise, then measure against it. Stops both premature optimisation and the endless tuning that never ships. Auto-triggers on: "make it faster", "is this fast enough", "optimize this", "performance budget", "it feels slow", "reduce latency"
ponytail
Make an approximation admit it is one. Find the heuristics, samples, regex parsers, caches and timeouts in a change and give each the one line saying what it gets wrong. Auto-triggers on: "is this exact", "add a caveat", "what are the limitations", "ponytail", "document the limits", "how accurate is this"
api-endpoint
Scaffold a new REST endpoint across the planes. Auto-triggers on: "add an endpoint", "new API route", "expose this over HTTP", "wire a handler"
write-test
Generate tests for existing code. Auto-triggers on: "write tests for", "add test coverage", "this needs tests"
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.