Browse reusable Agent Skills, each with a clear purpose and practical guidance.
完成した差分・PR・検証証拠に残る Unknown（未確認の前提・調査されていない影響・ 不足している証拠）を横断合成する evidence-sufficiency のメタ観点。個別 defect の 検出は既存 skill へ委譲し、本 skill は「そのリスク種別を調査した証拠が残っているか」 の meta 評価のみを行う。finding verification 後の合成ステップとして report-only で 実行し、残存 Unknown を output-format §4「Unverified / Residual Risk」の Unknown Coverage 下位構造へ、判定を既存 verdict 語彙（GO/ESCALATE/NO_GO）へ写像する。 新しい語彙・schema は作らない。
Use Chrome DevTools MCP for frontend performance work such as slow page loads, LCP, CLS, INP, Lighthouse regressions, scripting hotspots, or layout thrash. Trigger when the user asks for page speed analysis, Core Web Vitals debugging, trace-based investigation, or before-and-after performance verification.
Test-drive Django validation across model constraints, forms, ModelForms, views, template errors, duplicate handling, and redirects. Use when adding or refactoring Django forms, surfacing validation errors, deciding model-vs-form-vs-view responsibility, or preventing invalid input from being saved.
Drive Django features with outside-in and double-loop TDD: user story, Selenium or functional test, focused Django unit/integration tests, minimal implementation, and refactor. Use when starting or changing Django behavior with tests-first workflow, red/green/refactor, working-state-to-working-state refactors, or deciding which layer should own the next test.
Design, debug, and maintain Django Selenium functional tests with live server setup, explicit waits, test isolation, page objects, screenshots, CI artifacts, and staging checks. Use when browser tests are flaky, too slow, duplicated, failing in CI, need a page pattern, or should verify a Django user journey end to end.
Verify and monitor running experiments for operational quality. Use when designing prelaunch QA, spot-check tooling, experiment canaries, A/A tests, leakage checks, interference monitoring, active experiment dashboards, alerts, or an experimentation quality roadmap.
Detect flaky tests by running a test command N times
Plan evaluation strategies for machine-learning product changes. Use when deciding between offline evaluation, interleaving, online A/B tests, multi-armed bandits, or model filtering for ranking, recommendation, search, personalization, or other ML-powered user experiences.
Use this skill when the user needs Obsidian runtime debugging or developer introspection through official desktop Obsidian CLI developer commands, including devtools, console/errors capture, DOM/CSS inspection, screenshots, CDP calls, and controlled eval. Do not use for regular note CRUD, task/property editing, Sync/Publish administration, or community tooling outside the official CLI.