duongdev

Enforces quality verification gates with four-step validation (tests pass, build succeeds, checklist complete, no blockers) before task completion, PR creation, or status updates. Auto-activates when user says "done", "complete", "finished", "ready to merge", or runs /ccpm:verify or /ccpm:done commands. Provides systematic verification workflow that prevents false completion claims and ensures production readiness. Blocks external system writes (Jira, Slack) until evidence collected. Integrates with external-system-safety for confirmation workflow. When verification fails, suggests /ccpm:verify to debug issues systematically.

Guides optimal Linear operations usage with caching, performance patterns, and error handling. Auto-activates when implementing CCPM commands that interact with Linear. Prevents usage of non-existent Linear MCP tools.

Provides intelligent project setup and management with agent-based architecture to minimize token usage. Auto-activates when user mentions project setup, "add project", "configure project", "monorepo", "subdirectories", "switch project", or "project info". Uses three specialized agents internally: project-detector (detect active), project-config-loader (load settings with validation), project-context-manager (manage active project). Guides through four workflows: Add New Project (setup + templates), Configure Monorepo (pattern matching + subdirectories), Switch Between Projects (auto or manual), View Project Information. Provides templates for common architectures (fullstack-with-jira, fullstack-linear-only, mobile-app, monorepo). Validates configuration and suggests fixes for errors. Handles context-aware error handling with specific fix suggestions.

Systematic debugging with defense-in-depth approach (symptoms → immediate cause → root cause → systemic issues → prevention). Auto-activates when user mentions "error", "failing", "broken", "debug", "bug", "issue" or when /ccpm:verify runs. Uses Observe → Hypothesize → Test → Confirm → Fix → Verify → Document workflow. Updates Linear with findings and automatically logs blockers that require external fixes. Traces root causes instead of patching symptoms. Integrates with ccpm-code-review to verify fixes pass all gates before marking complete. Suggests binary search for intermittent issues and five-whys analysis for complex problems.

Discovers, manages, and troubleshoots MCP servers with three-tier classification (required: Linear/GitHub/Context7, optional: Jira/Confluence/Slack/BitBucket). Auto-activates when user asks "MCP server", "tools available", "Linear not working", "what tools do I have", or when plugin installation fails. Provides automatic server discovery, configuration validation, and health monitoring. Diagnoses connection issues (missing env vars, wrong config, network problems) with specific fix suggestions. Requires setup confirmation for optional PM integrations. Shows rate limit status and recommends optimizations when performance degrades.

Creates custom CCPM skills from request to deployment with proper templates, safety guardrails, and integration patterns. Auto-activates when user mentions "create skill", "custom workflow", "team specific", "extend CCPM", "codify team practice", or "reusable pattern". Guides through purpose definition (what skill does), activation triggers (when it runs), CCPM integration points, and safety rules. Provides three skill templates: Team Workflow (codify practices), Safety Enforcement (add checks), and Integration Skills (custom tools). Creates directory structure, frontmatter metadata, multi-phase instructions, and supporting docs. Tests skill activation before deployment and suggests improvements.

Provides conventional commits guidance and auto-generates commit messages from git changes. Integrates with /ccpm:commit for automated git commits linked to Linear issues. Auto-activates when users ask about committing, creating git commits, or discussing commit message formats.

Discovers and researches authoritative documentation with version-specific, context-aware search. Auto-activates when user asks "find documentation", "API docs", "how to use", "integration guide", "best practices", "design pattern", or when running /ccpm:plan or /ccpm:plan. Fetches latest docs from official sources via Context7 MCP. Uses progressive discovery (overview → API reference → integration → best practices). Prioritizes: Official docs → Framework guides → API references → Community resources. Provides version-specific recommendations and code examples from documentation. Flags important caveats and performance considerations. Surfaces migration guides when upgrading frameworks.

Use when working on CDP Browser's renderer (src/). Triggers on tasks involving React components, src/lib domain modules, Tailwind v4 + shadcn UI, Vitest tests, or anything visible in the Electron window. Loads the frontend conventions before code is written so the work follows established standards.

Scaffolds the next-numbered docs/adr/ file from the ADR template, filling the title, Status (Proposed), and Date lines. Use when the user says "write an ADR", "record this decision", "/adr", or an architectural decision needs recording.

Captures a raw, unstructured idea as a timestamped top entry in docs/memories/ideas.md. Use when the user says "capture an idea", "/idea", or floats a future improvement worth recording before it's lost.

Records a hard-won lesson as a timestamped top entry in docs/memories/learnings.md. Use when the user says "we learned", "note that", "/learn", or a non-obvious lesson surfaces that's worth remembering.

Scaffolds the next-numbered docs/tasks/ file from the task template, filling only the title line. Use when the user says "create a task", "new task", "/new-task", or a piece of work needs its own task file before it can be picked up.

Monitor a GitHub PR until approved — poll CI, read review comments, address blocking/should-fix findings, push fixes, re-poll. Use when user says "monitor the PR", "watch the PR", "address reviews", or "push until approved".

Appends a new open risk entry (next R-NNN, 🔴) to docs/memories/risks.md with a mitigation/trigger scaffold to fill in. Use when the user says "log a risk", runs /risk, or a new project risk surfaces in conversation.

Prints a one-screen orientation block — in-progress tasks, top open risks, next ready-to-pick tasks. Use when the user says "where are we", invokes /status, returns after a break, or asks what to work on next.

Render prose responses in bionic reading style by bolding the leading portion of each word, leaving code, paths, and commands untouched. Use when the user asks for bionic reading / bionic mode / bold-prefix formatting, or to toggle it off ("stop bionic", "normal formatting").

Analyze staged/unstaged changes and create a git commit that follows the project's detected conventions. Use when the user asks to commit, create a commit, or save changes to git. Push only if the user explicitly says "then push" in the same request.

Resolve git merge conflicts by semantically merging both sides of each conflict. Use when the user has merge conflicts, asks to resolve conflicts, or is stuck in a merge/rebase with conflict markers.

Review and freshen all documentation and notes in the current project — discover and obey the project's own doc conventions, fix drift, flag quality gaps, and fill genuinely missing context. Use when the user asks to update docs, freshen documentation, audit AGENTS.md/rules, or clean up notes.

Iteratively review and simplify uncommitted changes until clean or stuck. Use when the user asks to polish, clean up, or iteratively improve uncommitted code changes.

Run the full wrap-up pipeline — polish, then docs-revise, then a single commit — in one deliberate pass. Use ONLY when the user explicitly asks to run polish, docs-revise, and commit together as one pipeline (e.g. "/ship", "polish docs-revise and commit", "wrap this up and ship it"). Do NOT auto-trigger for a plain "clean this up", "commit", "polish", or "update docs" request — those are the individual skills.

Writing voice profile for AI-drafted communication. Provides channel-specific style samples and anti-patterns under profiles/ — pick the file matching the output destination (PR comment, chat, commit message, formal email). Use when the user says "draft in my voice", "rewrite this like I'd say it", or another skill needs a consistent writing register. This skill does not know about its consumers.

Cut a versioned release of the GLKVM Web Enhancements extension — bump the version in the two places that must match, regenerate the popup screenshot, ship, push, and publish a GitHub release. Use when the user asks to release, cut a release, publish a new version, or bump the extension version (e.g. "/release", "/release 1.1.0", "release a patch").