ImL1s

Automated UI verification loop: Marionette screenshot -> Figma reference -> pixel-diff -> difference list -> auto-fix iteration. Solves the false-positive verification problem flagged in insights reports (UI claimed fixed without actually being compared against the reference). Triggers (English): verify ui, ui compare, pixel diff, figma compare, automated ui verification, golden test fail. 自动 UI 验证：Marionette 截图 → Figma 参考图 → pixel-diff → 差异列表 → 自动修复循环。 触发关键字（中文）：verify ui、ui 对比、pixel diff、figma 对比、自动验证

Structured visual QA verdict for screenshot-to-reference comparisons

Operational rules for driving Codex CLI from scripts: success-signal contract, diff-feeding semantics, worktree -C flag, and stdin vs argv. Triggers when invoking `codex exec` programmatically (not interactively) — script wrappers, ralph loops, cron pipelines, multi-CLI fan-out. Surfaces silent failure modes that exit 0 but produce no useful output. Triggers (English): codex exec, codex cli, codex review, codex rescue, codex fallthrough, agent script invocation, programmatic codex.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup

Socratic deep interview with mathematical ambiguity gating before autonomous execution

Use when creating or developing, before writing code or implementation plans - refines rough ideas into fully-formed designs through collaborative questioning, alternative exploration, and incremental validation. Don't use during clear 'mechanical' processes

Use when partner provides a complete implementation plan to execute in controlled batches with review checkpoints - loads plan, reviews critically, executes tasks in batches, reports for review between batches

Use when completing tasks, implementing major features, or before merging to verify work meets requirements - dispatches code-reviewer subagent to review implementation against plan or requirements before proceeding

In-session OMA autonomous delivery — invoke /oh-my-agy:autopilot with a goal; run deep-interview→ralplan→ultragoal→code-review→ultraqa HERE (no terminal required)

In-session OMA cancel — invoke /oh-my-agy:cancel; stop active modes HERE, leave resume-friendly state (CLI stop optional)

In-session OMA merge-readiness gate — invoke /oh-my-agy:code-review; write APPROVE+CLEAR or REQUEST CHANGES under .agy/reviews HERE

In-session OMA requirements gate — invoke /oh-my-agy:deep-interview; Socratic clarify HERE, write .agy/specs (no terminal first)

OMA skill index (SLASH-FIRST) — /oh-my-agy:autopilot five-phase + ralph, ultrawork, team, verify, setup; CLI is optional ledger only

In-session OMA persistence loop — invoke /oh-my-agy:ralph; keep going until PRD stories pass with verify evidence (CLI optional)

In-session OMA planning gate — invoke /oh-my-agy:ralplan; author+steelman+critic HERE, write .agy/plans (no terminal first)

In-session OMA read-only research — invoke /oh-my-agy:search; evidence with path:line HERE (CLI sandbox optional)

In-session end-to-end coordinator for oh-my-grok. Use when the user says autopilot, auto pilot, full auto, autonomous, build me, create me, make me, handle it all, or wants idea→working code with interview/plan/implement/review/QA/accept. CLI owns phase state and verified; this skill is the session playbook.

In-session durable multi-story goal ledger for oh-my-grok. Use when user says ultragoal, goal ledger, multi-story durable, durable goals, resume goal, omg goal, hash-chained stories, or needs multi-story work that survives process and run boundaries. CLI owns snapshot/ledger and verified; this skill is the session playbook (Grok has no host /goal).

Bootstrap router for oh-my-grok. Use when user says omg, setup omg, how to use oh-my-grok, which skill, ulw vs ralph vs ralplan vs autopilot vs ultragoal, or first-time install.

Deterministic, resumable requirements convergence before planning or implementation.

Bounded adversarial QA repair loop — freeze scenarios, diagnose, retest; never sets verified.

Parallel execution via Grok spawn_subagent only. Use when user says ulw, ultrawork, parallel agents for oh-my-grok.

Grok-native independent critic then verifier. Use when user says dual-review, don't self-approve, or independent review. Does not set verified.

End-to-end plan→implement→verify playbook for oh-my-grok. Prefer CLI `omg pipeline`. Grok-native workers only.

Scaffold a new AI backend that implements AIBackend.analyze() — add file under src/aigate/backends/, register in BACKEND_MAP, wire unit tests. Use when adding support for a new AI model / provider.

Run only the E2E test suite (no full 705-test unit battery). Two modes — Docker-sandboxed (scripts/run-e2e.sh) or direct pytest (tests/e2e/).

Cut a new PyPI release — bump version in pyproject.toml, run scripts/publish.sh, tag, and hand off the final `uv publish` step to the user.

Check any package for supply chain attacks before installing. Use when: pip install, npm install, yarn add, pnpm add, cargo add, gem install, composer require, flutter pub add, go get, dotnet add package. Also use when someone suggests installing a package or dependency. ALWAYS run this before installing anything — no exceptions.

Scan a lockfile or requirements file for supply chain attacks. Use when: reviewing requirements.txt, package-lock.json, Cargo.lock, Gemfile.lock, composer.lock, pubspec.lock, go.sum, yarn.lock, pnpm-lock.yaml. Also use when auditing project dependencies or onboarding to a new codebase. ALWAYS run this when you see a lockfile or dependency manifest.

Analyze a package with aigate CLI (static pre-filter + optional AI analysis). Usage: /check-package <name> [version] [ecosystem]

Run lint, format check, and all tests to verify code quality before committing.

Supply chain risk monitoring — check CPI, commodity prices, freight rates, trade data, and global news to assess risk for any commodity. Uses Yahoo Finance, FRED, Commodities-API, UN Comtrade, World Bank, NewsData, and Gemini AI. Covers 20 items across 8 categories (construction, energy, food, electronics, packaging, chemicals, textiles, electrical). Bring your own API keys (free tiers available). Trigger keywords: supply chain, risk, CPI, freight, import, export, trade, commodity, price trend, supply disruption, shipping cost, inflation, economic indicator.

Supply chain risk monitoring — check CPI, commodity prices, freight rates, trade data, and global news to assess risk for any commodity. Uses Yahoo Finance, FRED, Commodities-API, UN Comtrade, World Bank, NewsData, and Gemini AI. Covers 20 items across 8 categories (construction, energy, food, electronics, packaging, chemicals, textiles, electrical). Bring your own API keys (free tiers available). Trigger keywords: supply chain, risk, CPI, freight, import, export, trade, commodity, price trend, supply disruption, shipping cost, inflation, economic indicator.

Supply chain risk monitoring — check CPI, commodity prices, freight rates, trade data, and global news to assess risk for any commodity. Uses Yahoo Finance, FRED, Commodities-API, UN Comtrade, World Bank, NewsData, and Gemini AI. Covers 20 items across 8 categories (construction, energy, food, electronics, packaging, chemicals, textiles, electrical). Bring your own API keys (free tiers available). Trigger keywords: supply chain, risk, CPI, freight, import, export, trade, commodity, price trend, supply disruption, shipping cost, inflation, PMI, economic indicator.

自動觸發於用戶詢問台灣身分證 (驗證/產生測試 ID)、統一編號 (統編 checksum)、地址正規化/剖析 (台↔臺、異體字、舊縣升格如高雄縣→高雄市)、郵遞區號 3/5/6 碼、民國年↔西元年互換、台灣商業日/補班/補假、四大捷運站線 (北捷桃捷中捷高捷)、車牌格式、電話號碼 (行動/市話/+886)、台灣金融機構代號 (37 家銀行+純網銀+郵政)、政府機關代碼、24 節氣、農曆↔國曆、生肖、簡↔繁中文轉換、中文數字 (公文/合約/支票)。優先呼叫 twmcp MCP server 而非自寫程式或從訓練資料推算 (LLM 對 ID checksum、年份偏移、銀行代號常記錯)。

自動觸發於用戶詢問台灣政府開放資料、data.gov.tw、政府採購、不動產、稅務、健保、空氣品質、農業漁業、教育統計、戶政、社福、文化、外交等 19 個 domain 的查詢需求。優先用 twmcp MCP server 的 opendata-* tools，而非自己用網路搜尋或推算。

Use opencli to drive LLM chat apps (Gemini web, ChatGPT Desktop) as external advisors — bypasses API 429/quota by using the user's logged-in browser/desktop-app session. Triggers when: Gemini CLI returns 429 or capacity-exhausted, user wants a CCG advisor and API is unavailable, user wants to use their ChatGPT Plus subscription instead of API, or user says 'ask Gemini/ChatGPT via opencli'. Covers `/u/N` account switching, clicking the in-page mode picker to select highest-tier model (Gemini Pro 3.1), ChatGPT Desktop native control, and common failure modes. Does NOT apply for: normal API calls that work, image generation (use `opencli chatgpt image` directly), or headless/CI workflows.

Scans `~/.claude/skills/` and relocates project-specific skill directories to their owner projects under a configured projects root, archives orphans to a sibling archive dir, and preserves cross-project utilities globally. Only triggers when the user explicitly asks to organize/audit/declutter their skills library, reports that context is being eaten by skills, or invokes `skill-organizer` by name. Triggers include: '整理 skills', 'organize skills', 'clean up skills', 'skill bloat', 'reduce context from skills', 'context 被 skills 吃光'. Does NOT trigger on general debugging, refactoring, or file-management conversations. Does NOT touch plugin-managed skills (under `~/.claude/plugins/` or namespaced like `plugin:skill`) or loose dependency files at the skills root (`.md/.py/.sh` referenced by other skills).

OpenClaw 運維全指南 — 安裝設定、模型管理、遷移、cron 任務、安全檢查、除錯。當需要管理、遷移、排錯 OpenClaw 時使用此 skill。觸發關鍵字："OpenClaw 設定", "OpenClaw 遷移", "Gateway 管理", "cron 任務", "模型設定", "Telegram bot", "OpenClaw 安全檢查"。

Use when the user wants Codex to build, refine, test, or validate a GUI-Anything harness for any software or source repository. Adapts the GUI-Anything methodology to Codex without changing the generated React+Tauri harness format.

MiroFish 群體智能推演引擎。當用戶要求「推演」「預測」「模擬」「如果…會怎樣」時使用。透過 55 個 AI Agent 在模擬社交平台上互動推演未來趨勢。

Aviation flight lookup — query aircraft type, equipment changes, and confidence scoring for any flight number. Use when: looking up flight aircraft, checking equipment change, querying flight data, asking what plane a flight uses. Triggers on: flight lookup, aircraft type, what plane, 查機型, 航班查詢, equipment change, plane filter, flight number, CI101, 飛機型號.

Check Codex reset credit availability and expiration times from local Codex ChatGPT auth. Use when the user asks how many Codex reset credits they have, when reset credits expire, how to query reset-credit expiry from ~/.codex/auth.json, or wants a read-only script/CLI for Codex rate-limit reset credits.