evaluate-zephyr-perf
Run a perf gate on a PR that touches lib/zephyr internals.
Browse reusable Agent Skills, each with a clear purpose and practical guidance.
Run a perf gate on a PR that touches lib/zephyr internals.
In CI, run the infra/lint catalog review over a PR.
Profile JAX training and analyze hotspots. Use when profiling or optimizing training throughput.
Diagnose a bug systematically instead of guessing — reproduce, isolate, form hypotheses, and test them to root cause. Use when debugging, chasing a defect, an intermittent failure, or 'why is this happening?'. Produces a structured diagnosis: a reliable repro, the narrowed-down location, ranked hypotheses with how to test each, and the root cause + fix once found.
Plan tests for an API endpoint or service — functional, negative, and contract. Use when asked to test an API, write API test cases, plan REST/GraphQL endpoint testing, or validate an API contract. Produces an API test plan — per-endpoint cases (status codes, schema, auth, validation, errors), boundary/negative cases, contract checks, and non-functional notes — so the API is verified beyond the happy 200.
Write a clear, reproducible bug report that gets fixed fast. Use when asked to write a bug report, file a defect, report an issue, or turn 'it's broken' into an actionable ticket. Produces a structured report — a precise title, steps to reproduce, expected vs. actual, environment, severity/priority, and evidence — so a developer can reproduce and fix it without a back-and-forth.
Review a product/codebase for internationalization readiness before you localize. Use when asked if a product is ready to localize, to review i18n readiness, find hard-coded strings/locale bugs, or prep for going multilingual. Produces a readiness audit — externalized strings, locale-aware formatting, layout/expansion, encoding/RTL, and a prioritised list of i18n fixes to make before translation starts.
Drive a feature with a disciplined test-driven development loop — red, green, refactor. Use when implementing a feature or fixing a bug and you want tests to lead, or when asked to 'do this with TDD' / write the test first. Produces a step-by-step red-green-refactor plan: the failing test to write first, the minimal code to pass it, and the refactor — one small cycle at a time.
bug-hunter 阶段 1 技能。负责提取代码改动、执行敏感信息脱敏,并按文件/代码块生成多轮随机化输入以缓解 LLM 位置偏差。