Explore

Use when the user is debugging a bug, tracing an error, or asking why something fails. Examples: "Why is X failing?", "Where does this error come from?", "Trace this bug"

Use when the user needs to run GitNexus CLI commands like analyze/index a repo, check status, clean the index, generate a wiki, or list indexed repos. Examples: "Index this repo", "Reanalyze the codebase", "Generate a wiki"

Use to check ONE already-written piece of copy against a voice guide before it ships — it spawns ONE fresh sibling session via CCC that scores the existing draft against the stated voice rules, flags every violation with the offending quote and a minimal in-voice rewrite, and gives a ship / revise verdict. For an artifact that already exists, not a copy competition.

Use before shipping a change that touches a trust boundary (a new endpoint, input parser, file-path handler, subprocess call, or auth check) — it spawns ONE fresh sibling session via CCC that maps the abuse surface of that specific diff, ranks each abuse case by impact and likelihood, and hands back a prioritized ledger that feeds /security-review. A defensive threat map, not an exploit run.

Use when a superpowers plan (writing-plans / brainstorming output saved locally) is ready and you want durable, dashboard-visible tracking instead of an in-session scratch ledger — it imports the plan into a Watchtower queue as tickets, then optionally dispatches one CCC lane per ticket that closes its ticket with a summary. Bridges superpowers (ephemeral subagents, file-tracked) to Watchtower (durable queue) to CCC (fleet dashboard).

Use when you want a one-line status from every live sibling session in a repo (or across repos) collated into a standup digest — queries siblings via CCC's `/api/ask` and spawns nothing.

Spawn a repo-scoped UX-fixes worker session that drains only one repo's UX-fixes queue, via the CCC API.

Use when, before committing to a non-trivial approach, diagnosis, or answer, you want an independent check — it spawns ONE fresh sibling session via CCC (no shared context) to solve the same task from scratch, then diffs its answer against yours. Disagreement is the product.

Use right before cutting a release — it spawns ONE fresh sibling session via CCC that audits only the JUDGMENT gates a release script cannot check (does the changelog actually describe the diff, does the version bump match semver for what changed, do the docs mention the new user-visible feature, is anything uncommitted or half-shipped) and returns a GO / NO-GO with blockers. Advisory only; it never touches the release.

Use when one release needs to go out across several channels at once — spawns one writer sibling session per channel via CCC (changelog, LinkedIn, X thread, optional blog) off a single source pack, then collates the drafts.

Use after you fix a bug and before you claim it fixed — it spawns ONE skeptic sibling session via CCC that must first reproduce the ORIGINAL bug on a clean worktree at the pre-fix ref, then prove your fix actually moves it. Reproduce-first discipline, independent of your context.

Check CCC's cross-session memory before building a feature or acting on "earlier work" claims.

Coordinate parallel sessions for discussion, task execution, and git commits.

Use after a UI or user-visible change and before you claim it works — it spawns ONE CCC verification lane that drives a real browser (gstack browse, or CCC's puppeteer snapshot as fallback) against the running app, checks the specific thing you changed, and reports a visual/behavioral verdict with a screenshot. Independent of your context; catches what a diff read and a unit test cannot.

Use when dispatching a batch of autonomous CCC worker lanes with a mission brief and a report_to return address, or when receiving a lane assignment and needing to send a well-formed completion report — and, on the dispatcher side, before trusting a lane's SUCCEEDED report at face value.

Use to catch documentation that has silently gone stale — it spawns ONE fresh sibling session via CCC that extracts every checkable factual claim from a doc (API fields, endpoint names, CLI flags, config keys, file paths, version numbers, example commands) and verifies each one against the actual code, reporting each mismatch as a drift with severity. A claim-by-claim audit, not a read-through.

Spawn, inject into, and ask questions of persistent sibling sessions via Claude Command Center (CCC).

Use when a bug has already survived at least one debugging attempt and you have 2-3 genuinely distinct root-cause hypotheses — spawns one racer sibling session per hypothesis via CCC, referees the first CONFIRMED root cause with reproducible evidence, and stands the others down via inject.

Use when you want a marketing-copy A/B — spawns two writer sibling sessions via CCC that draft the SAME message architecture in DIFFERENT voices, then a judge session (which saw neither draft happen) scores both and picks a winner.

Use when the user asks to open/show the CCC throughput (token) dashboard, or asks how many tokens they've used — overall/so far, in the last 7 days, or for a specific session (e.g. "how many tokens is session ffbbeef8", "open throughput dashboard", "what's my token burn"). Answers token/cost questions from CCC's throughput API and links the page.

How to bring up and runtime-test the CCC dashboard (server.py) on a Linux box — ports, duplicate-instance bypass, fresh-install simulation, forcing best-effort/[ccc:swallowed] failures, and hook stdin tests.

Use when reconciling existing GitHub PRs, issues, listings, or outreach-tracker targets for codex-profiles after submission or review. Not for lead discovery, qualification, drafting, or outreach.

Use when qualifying outreach-tracker GitHub lead candidates for codex-profiles after lead generation and before any closing draft. Not for discovery, drafting, outreach, submission, or monitoring.

Use when running GitHub lead generation for codex-profiles. Finds repository candidates and performs shallow tracker intake before lead qualification. Not for drafting, outreach, or monitoring.

Use when drafting GitHub PRs, issues, listing copy, or maintainer requests for qualified codex-profiles tracker targets. Not for discovery, qualification, monitoring, outreach, or submission.

Produce publish-ready Chinese and English writing packages across WeChat articles, Medium/Substack essays, research notes, founder memos, op-eds, speeches, data storytelling, statistical graphics, and platform-ready HTML. Use this unified workflow for writing, rewriting, sourcing, lowering AI feel, creating visuals, and packaging final articles for Claude Code, Claude CLI, Codex CLI, and local agents.

Keeps an agent building and validating real output instead of servicing its own bookkeeping. Use for multi-stage pipelines, capability roadmaps, long-running implementation loops, migrations, staged data or build runs, and any run where administrative hashes, locks, receipts, dashboards, certification markers, or progress metadata can block correct work. Use when an agent refuses to advance, rewinds finished stages, or reruns unchanged work because a marker is missing or stale.

Search local coding-agent session history and identify safe resume commands. Use when the user asks to find, inspect, continue, or recover previous work from Claude Code, Codex, Pi, or another agent indexed by fast-resume.

生成一份对外可分享、脱敏的 AI-Native 开发者 README。 量化展示我对 Claude Code + Codex CLI + Kiro (AWS) + Trae (ByteDance) + Gemini Antigravity (Google) + Cursor 的使用深度、AI 协作风格、 项目与领域分布、兴趣主题，以及与 GitHub 提交的产出关联。 Trigger when the user says: "生成我的 AI 档案" / "做一份 AI-native README" / "分析我的 Claude / Codex / Kiro / Trae / Antigravity / Cursor 使用情况" / "总结我的 AI 使用" / "生成 AI 月度报告" / "按月份分析我的 AI 编码" / "分析 2026-05 的 AI 使用" / "build my AI usage profile" / "build my monthly AI coding report" / "analyze my AI usage for May 2026" / "summarize my Claude / Codex / Kiro / Trae / Antigravity / Cursor history" / "生成开发者画像". 全程本地、只读、默认匿名、不上传任何数据。

View and configure settings for coding agents (Claude Code, Codex CLI, OpenCode, Devin CLI/Desktop, and others). Covers JSON settings for Claude Code and Devin, TOML for Codex CLI, and JSON/JSONC for OpenCode, including permissions, sandbox, model selection, profiles, feature flags, providers, hooks, subagents, and skills.

Rename and refactor C# symbols in a .NET solution or multi-solution monorepo with a one-shot Roslyn CLI. Use when the user asks to rename a symbol, preview impact, update references across a solution, or refactor shared projects across several solutions.

Use this skill when the user asks to plan, design, scope, estimate, or implement a feature, bug fix, refactor, migration, integration, API change, UI change, or other project modification. Enforces a planning gate before editing code — investigate project context, analyze the task, surface ambiguities, contradictions, risks, dependencies, and blockers, ask focused questions, produce an evidence-based step-by-step plan, and implement only after explicit user approval. Not for trivial one-line edits, pure questions about the codebase, or changes the user has already reviewed and approved for direct implementation.

Search, install, configure, update, and remove MCP servers across coding agents (Claude Code, Cursor, VS Code, Claude Desktop, Gemini CLI, Codex, Goose, Zed, Devin CLI/Desktop, and more). Supports multi-agent installation via npx add-mcp, the official MCP registry, and direct config editing.

Install, upgrade, configure, and verify developer CLI tools safely. Use when a user asks to install a new CLI, command-line app, SDK tool, package-manager binary, GitHub release binary, language runtime tool, or AI/vendor CLI; configure shell PATH/completions; run first login; set API keys, tokens, or env variables for a CLI; migrate an existing CLI install; or troubleshoot a CLI installation while avoiding secret leakage.

Copy text to clipboard with optional rich formatting. Triggers on "copy to clipboard", "copy that", "pbcopy", "copy formatted", "copy rich text".

8-step disciplined bug-fix protocol that treats every production bug as two failures — the code defect itself and the testing system that allowed it through. Use when fixing a production bug, investigating a regression, writing a post-mortem, or auditing a missed defect. Triggers on "fix this bug", "production bug", "regression test", "post-mortem", "test gap", "why did the tests miss this".

Decision-frontier wayfinding: build and work a durable map of open design decisions BEFORE planning, for large or foggy efforts where $map-plan would force premature decomposition. Use when a task is too big or too vague to decompose — many unknowns, tangled decisions, or "I'm not even sure what to build yet" — and you want to resolve the key decisions one at a time (research, prototype, grilling, task tickets) behind a claim-before-work frontier, with a "fog of war" for questions you cannot yet state sharply. Produces a handoff that $map-plan consumes. Modes are explicit: chart (start a map), work (resolve one ticket), handoff (finish). Do NOT use when scope is already clear enough to specify — go straight to $map-plan; do NOT use to implement code or to decompose an already-specifiable change.

Interactive deep-understanding and quiz mode for MAP sessions. Use when the user wants to understand code, a diff, workflow result, debugging cause, or architecture and be checked with restatements or quizzes.

Show per-subtask/agent token accounting (input, output, cache read/creation, cost, cache-hit ratio) for the current branch. Supports --dashboard (box-drawing visual), --history (trends), --estimate (cost projection), --json / --csv export, and --finalize (record session snapshot). Use when asked for token usage, run cost, or a token report. Do NOT use to plan or run work; use map-efficient.

TDD MAP workflow: write tests from the spec FIRST, then implement, so tests validate intent not implementation. Use when correctness is critical (auth, payments, data integrity). Do NOT use without a spec; use map-efficient instead.

Execute a single subtask from an existing MAP plan via Actor and Monitor. Use when map-plan has decomposed work and you want fine-grained control over one subtask. Do NOT use without an existing plan; run map-plan first.

Evaluate a $map-* skill's trigger accuracy and cost. Use when asked to measure skill trigger accuracy, run an eval-set, or check token/duration cost via `mapify skill-eval`. Do NOT use to plan or implement; use map-plan or map-efficient.

Interactive 4-section code review using Monitor, Predictor, and Evaluator agents plus the user and maintainer role reviewers on current changes. Use when reviewing a diff, PR, or staged work before merge. Do NOT use to plan or implement; use map-plan or map-efficient.

Execute the mapify-cli package release workflow with validation gates and PyPI publication. Use when shipping a new MAP Framework release. Do NOT use for ordinary feature work; use map-efficient.

Use when reviewing a PRD, product brief, feature brief, or requirements document before planning or engineering handoff. Produces an evidence-backed 0-10 readiness score, strengths, weaknesses/risks, and uncovered edge cases across 13 dimensions. Do NOT use as a substitute for $map-plan, for code review, or for tiny engineering tasks with no PRD.

ARCHITECT phase - decompose complex tasks into atomic subtasks with research, spec, and branch-scoped plan artifacts under .map.

Finalize cross-session memory on demand; --finalize-all sweeps every dirty scratch. Use when ending a long session or before switching branches. Do NOT use for routine edits — finalize auto-runs at next SessionStart. Requires Codex CLI + git.

Capture reusable lessons after a completed MAP workflow. Use when a MAP run has finished and you want audited rules written to `.map/learned/` and promoted into the managed MAP section of `AGENTS.md`. Do NOT use during active implementation.

Minimal MAP workflow for small low-risk changes (40-50% token savings, no Predictor/Reflector). Use when the change is small, low-risk, and learning is not needed. Do NOT use for risky or complex work; use map-efficient.

Explain code, a diff, or the whole project the way a knowledgeable colleague would — result and main scenario first, then participants, flow, mechanisms, rules and constraints in progressive depth, with ASCII diagrams; for a PR or diff, exactly the change against the base with new behavior separated from existing context, documentation clarifications and unimplemented plans. Use when learning unfamiliar code, onboarding to a system, or auditing what a PR really does.

State-machine MAP execution workflow for Codex. Use when implementing an approved MAP plan end to end, resuming from branch MAP task_plan or step_state.json artifacts, or running non-trivial multi-subtask work. Use map-fast for tiny one-shot edits.

Structured MAP debugging via decomposer, actor, and monitor agents. Use when reproducing a bug, isolating a regression, or diagnosing an error with specialized agents — including failing or flaky tests (pytest AssertionError), crashes and segmentation faults, memory-corruption or memory errors in native/C extensions, intermittent or load-dependent failures (e.g. 500s under load), data-corruption bugs that only appear in production, scripts or hooks that silently exit or produce no output, and any "find the root cause" / "walk me through diagnosing" / "help me investigate" request. Trigger on phrasing like "failing test", "mysterious error", "segfault", "memory error", "intermittently fails", "root cause", "isolate the cause", "debug why", "diagnose", or "investigate the error". Prefer this over generic investigation when the user wants a systematic decompose-reproduce-fix-verify workflow. Do NOT use for greenfield features; use map-plan or map-efficient.

Run quality gates (lint, types, tests) and verify MAP workflow completion. Use when user asks to run checks, validate a workflow, or confirm a MAP run is done. Do NOT use to plan or execute new tasks; use map-plan or map-efficient.

Audit local AI coding-agent sessions with agenttrace. Use when the user asks to inspect Claude Code, Codex CLI, Qwen Code, Cline, Aider, Cursor exports, Hermes Agent, OpenCode, OpenClaw, Pi, Oh My Pi, Kimi CLI, Copilot-style logs, or generic JSON/JSONL traces for cost, tokens, tool failures, latency, anomalies, health, diffs, or CI gates.

Shape output for a reader with ADHD, then write each sentence in controlled English: action first, one word one meaning, active voice, simple tenses, state restated every turn. Invoke with /attention-control; stays on until "stop attention control".

Minimize token consumption & maximize prompt cache hit rate. Use when user asks to save tokens, reduce cost, improve cache hit rate, or be more concise.

Persistent global memory across conversations. Read memory file at start of every conversation, write important context during conversation so it survives context compression. Use ALWAYS. Triggered automatically.

Cuts a new versioned release of codex-trace end-to-end without asking any questions. Detects commits since the last ancestral tag, classifies them with conventional-commit rules to pick the semver bump, bumps every version-bearing file in sync, writes the CHANGELOG entry, commits and tags on a release branch, pushes the tag to trigger the GitHub Actions release pipeline, fast-forwards `main` onto the release commit and pushes it, then cleans up the local branch. Use this skill whenever the user mentions cutting, tagging, bumping, or shipping a release — including "cut a release", "release v1.2.3", "tag and release", "bump version", "publish a release", "do a patch release", or "make a new release". Trigger even when the user does not name the version — the skill computes it. Also use proactively when the user finishes a unit of work and asks to publish it.

Explain what just happened under the hood — the context, MCP tools, and hooks behind Dex's last response — as AI education. Use when the user says 'how did you do that', 'what just happened', 'explain the mechanics'. Not for a system health check; use `dex-doctor`.

Re-run the setup questions over an existing Dex vault when preferences or setup answers changed, without losing data. Use when the user says 'restructure my Dex', 'redo my setup', 'my pillars are wrong now'. Not for a job or role change; use `change-job` — it runs this same reset plus the people, archive and task passes. Not for first-time setup; use `setup`. Not for just toggling one feature; use `manage-capabilities`.