kc-claude-plugins

Use when reviewing a GitHub PR and posting inline review comments. Triggered by PR number, PR URL, 'review this PR', 'review current branch PR'.

Use when compiling E2E flow YAML to standalone bash test scripts. Triggers on "/e2e-compile", "compile flow", "recompile flow".

Use when generating, verifying, or smoke-testing E2E flow YAML from plans, specs, PRs, or mapped pages. Supports browser and CLI-only (Execute external) flows. Triggers on "e2e flow", "generate flow", "verify flow", "smoke test", "cli flow".

Use when asked to 'gemini review', 'gemini challenge', 'ask gemini', or want an independent cross-model second opinion / adversarial review / consultation from a different AI vendor (Google's Antigravity CLI, `agy`) on a diff, plan, or codebase question.

Use when a GitHub PR has review comments or automated feedback that needs triage, validation, and structured response before pushing fixes.

Use when reviewing a teammate's plan, Linear project/issue, design doc, or Slack-thread proposal AS the EM before build starts — applies Kent's EM lenses, verifies the single load-bearing assumption against real code, de-dups against existing reviews, runs a mandatory cross-model pressure-test, and delivers a calibrated, venue-ready message. Triggers: "review this plan/project as EM", "EM review", "plan review", or a Linear project/issue URL paired with "review".

Use when user says 'handoff', 'prepare handoff', '準備交接', or when context pressure is high and work state needs preservation before session ends.

Use when user says 'create pr', 'open pr', '建立 PR', '開 PR', '發 PR', '送審', or when implementation is complete and changes need a pull request on GitHub.

Use when creating a new Claude Code plugin, auditing an existing plugin for quality, improving plugin skills with TDD, or verifying plugin agent configurations.

Use before publishing a public plugin (especially before /kc-marketplace-sync) to grep for known leak patterns in the plugin's public reference files. Triggered by "/kc-plugin-forge sanitize-check <plugin>", "sanitize-check", "check leaks before publish", "prepublish scan". Read-only safety net — NOT a replacement for Early-stage Dreaming (which is the real privacy gate at capture time).

Use when presenting 2+ technical approaches, listing pros/cons, evaluating architecture or technology trade-offs, or when the user asks 'which should I choose'.

Use when triaging Linear issues as EM — exploring codebase, researching tech feasibility, decomposing oversized issues, posting structured implementation guidance, or processing meeting notes into Linear Customer Requests/Needs.

Use when verifying that a fix actually reaches the bug's break-point in the real runtime path — not just at unit-test level. Primary trigger is bugfix / cross-stack PR review (invoked by kc-pr-review). Also invoke standalone when pressure-testing a fix claim, pre-merge verification of cross-layer wiring, or debugging "why is this still broken". Output is a strict YAML contract declaring precision level (A/B/C/D), verified vs unverified failure-chain steps, residual uncertainty, and recommended human probes. Designed to prevent silent verification failure where an agent reports "verified" without actually exercising the runtime path.

Use when a feature is complete with PR and demo artifacts (e2e walkthroughs, Loom, screenshots), and user wants to announce on Slack. Triggers on "announce", "post to product", "draft product message", "公告", or after PR + demo completion.

Use when user says 'start daemon', 'stop daemon', 'daemon status', 'pr daemon', 'daemon config', 'daemon 掛了', 'daemon error', '啟動 daemon', '停止 daemon', 'daemon 狀態', or wants to manage, configure, or troubleshoot the PR review closed-loop daemon.

Use when a feature branch has messy interleaved commits that need reorganization into logical groups before PR review. Triggers include "squash commits", "clean up history", "reorganize commits", "reorder commits", "too many commits". Not for simple squash-all — use git merge --squash instead.

Use when syncing team context (initiatives, projects, members, cycles, customers, estimate scale) from Linear to local YAML cache. Triggers on '/kc-em-sync', 'sync team context', 'refresh team data', '同步團隊資料', 'team sync'.

This skill should be used when the user asks to "update project status", "write pulse update", "draft weekly update", "post project pulse", "project 更新", "寫 pulse", "weekly update", or wants to draft and post Linear project status updates. Covers data collection, health assessment, drafting, and posting.

Use when verifying static UI differences — computed CSS values on specific DOM elements, design tokens, theme changes, layout dimensions, pseudo-element styles. Declarative YAML input, browser-driven, pass/fail per check. Complements e2e-flow (which verifies dynamic behavior). Triggers on "ui verify", "check ui tokens", "computed style regression", "design token check".

Use when debugging e2e-test/e2e-map/e2e-walkthrough failures, maintaining these skills, or evaluating execution results. Triggers on "fix e2e skill", "e2e skill broke", "e2e gap analysis".

Use when running browser E2E tests — executing flow files, batch-testing across sites, or recording runs with video. Triggers on "run e2e", "e2e test", "browser test", "test with video".

Use when walking through UI flows interactively with human guidance — visual QA, exploring features, or recording demos. Records video by default. Triggers on "walkthrough", "walk the UI", "visual QA", "demo recording".

Use when debugging frontend runtime bugs by injecting console.log, observing in browser, and auto-cleaning up. Triggers on "e2e debug", "debug in browser", "inject logs", "runtime debug".

Use when the user invokes /e2e-dispatch or asks for the E2E operations menu. Router — prefer e2e-test, e2e-map, e2e-walkthrough, e2e-skill-ops directly for specific operations.

Use when syncing E2E plugin documentation after skill/agent changes. Diff-aware scan, history enrichment, and live probe verification. Triggers on "e2e doc sync", "/e2e-doc-sync", "check doc coverage".

Use when users need help discovering E2E pipeline commands or want to report doc gaps. Triggers on "e2e help", "/e2e-help", "how do I e2e".

Use when creating or updating UI element mappings for E2E testing — capturing app structure or refreshing stale mappings. Triggers on "e2e map", "map the UI", "update mapping".

Use when user says 'resume', 'continue', '繼續', 'pick up where I left off', or mentions a branch name to resume previous work. Triggers on session handoff resume prompts.

Manually cache an insight about a file or module to the context lake. Use when you want to preserve understanding about code you just explored. Also shows lake status and metrics.

Use when the user wants to set up or check their Claude Code statusline with Anthropic usage quota display. Triggers on "setup statusline", "statusline", "show usage", "quota display", "安裝 statusline", "statusline 設定". NOT for direct usage display — this skill installs the statusline component.

Use when running nightly plugin improvement cycle — forge validation, signal harvesting, north-star gap analysis, auto-fix and proposal generation. Triggered by cron or manual invocation.

Use when plugin documentation needs updating after feature changes, when verifying existing docs accurately reflect actual skill/agent behavior, or for periodic documentation maintenance. Scans source code, enriches with usage history from journal and memory, writes or updates docs, then reports gaps.

Use when users need help with Plugin Forge commands, want to discover available routes, or report documentation gaps. Triggers on "forge help", "how do I forge", "what forge commands", "kc-plugin-forge-help", "help with plugin forge", "what can forge do", "plugin forge help".

Use when setting up, configuring, or managing kc-nightwatch scheduling and settings. Triggered by '/kc-nightwatch-config', or when user asks to schedule nightwatch, change cron time, configure Slack channel, or check what plugins are monitored.

Use when checking nightwatch status, viewing last run results, reviewing improvement history, or reviewing pending proposals. Triggered by '/kc-nightwatch-report', '/kc-nightwatch-report --last', '/kc-nightwatch-report --review', or asking about nightwatch activity.