菜单
SkillsMP 已收集 CleanExpo/skills-library 中的 150 个 Skill。打开任一 Skill 可查看来源和详情。
已展示 40 / 150 个已收集 Skill。
Dispatch the eight operational roles against a task, each grounded in the vault before it reads any code, with a live view of what is running. Use for work that wants several independent lenses at once — map, plan, build, verify, review, security, CI, release…
原文语言：英语
Thin wrapper around the Pi-CEO Hour-1 provisioner. Given a client slug + email, invokes `swarm/inbox/provisioner.py` to spin up the full Linear project + Supabase portal_content + ContextBot Telegram intake bot + welcome email + Phill ping. Use when the user…
原文语言：英语
PM-Core — the first autonomous execution agent. Claims the highest-priority agent-ready ticket from the Unite-Group Linear board, clones the repo, implements the fix on a feature branch, runs CI, and opens a PR. Model: claude-sonnet-5 (senior agent tier).…
原文语言：英语
Full SEO toolkit for Claude Code, powered by the DataForSEO API. Replaces the data layer of an SEO agency: keyword research, competitor analysis, rank tracking, backlink audits, technical site crawls, content gap analysis, and client-ready PDF reports — all…
原文语言：英语
Use before any git push, pull-request creation or ready-for-review action, including autonomous loops and follow-up fixes.
原文语言：英语
Dictate a goal and ramble at it. Grounds in the project's North Star, structures the intent, researches, drafts a build-ready brief, then has Codex challenge it until it holds.
原文语言：英语
Use when fleet machines have stale/diverged skills or config ("update didn't reach the other computer"), when setting up a new machine, or when the estate-sync log shows CONFLICT/FAIL. Owns the automatic two-way propagation of the skills-library hub…
原文语言：英语
Run every gate CI's "Quality Checks" job runs, LOCALLY, before you push or open a PR — so a locally-green branch never opens UNSTABLE. Use before any git push / gh pr create in RestoreAssist, and whenever a PR shows mergeStateStatus UNSTABLE with a red…
原文语言：英语
Turns any goal into one short, paste-ready "gauntlet loop" prompt - a prompt that makes an agent set a concrete quality bar, split the work into small judgeable pieces, run a builder and a separate harsh critic on each, compare blind against the bar, and loop…
原文语言：英语
Emits a matched pair of gauntlet-loop prompts - one for the Claude Desktop app, one for the Claude CLI - so both grind separate projects at once without colliding, each looping until the board holds 100% AAA+. Dispatched by gauntlet-loop when the user wants…
原文语言：英语
The gauntlet variant for getting a real product to market - grinds a ship backlog to live instead of grinding one surface against a competitor. The bar is a production-observable gate per item, not a taste comparison; selection is by ship-blocking rank, not…
原文语言：英语
Use whenever about to mark a test, gate, verification or task as passing / GREEN / complete / fixed / shipped — the moment a CLAIM is made, not when a check is written. Catches vacuous verification, claims asserted but never checked against the live system,…
原文语言：英语
Use to curate machine-local memory from recent session transcripts — invoked as /dream, /dream status, or /dream apply 1,3. Reads ~/.claude/projects/**/*.jsonl from the last 24h, dedupes by session id, excludes the current session and sessions that produced…
原文语言：英语
Durable session handoff (/session-handoff). Gates the tree green via scripts/handoff-loop.sh, classifies the session SHIPPED / READY-TO-SHIP / WIP-BLOCKED against a Definition-of-Done, then generates a precise handoff before stopping, switching terminals,…
原文语言：英语
AI Communication Intelligence layer — the bidirectional cross-specialist consultation protocol that makes specialist agents TALK TO EACH OTHER before an artifact is finalized. Use when a task spans domains and the output needs input from multiple specialists…
原文语言：英语
Use BEFORE writing or changing any check, test, gate, guard or assertion — while the control is being built. Answers ONE question, will this fire when the defect is present. Catches controls that cannot fail, negative controls that planted nothing, probes…
原文语言：英语
Use when a secret, token, password or key is about to be written into a test, handled during a verification, or read at startup. Answers ONE question, does anything have to TOUCH the secret that should not. Catches fixtures that write literal secret-shaped…
原文语言：英语
Discipline for any branch that ends in a reviewed, gated, or protected PR. Load when starting or resuming such a branch, when a PR will not merge, when a bot reviewer (e.g. CodeRabbit) posts threads, when entering a fix-review loop with an independent…
原文语言：英语
Use when a check has RUN and you are about to say what it proved, or when choosing where to plant a canary, narrow a search, or enumerate the surface a guard covers. Answers ONE question, what did the instrument actually look at and what sentence am I…
原文语言：英语
Use when about to INTERPRET the result of a check you did not write — a CI conclusion, a test verdict, a gate outcome, a health payload, a query returning zero rows — and above all when about to dismiss a failure as pre-existing, environmental, flaky,…
原文语言：英语
Review a proposed fix with a different model in a fresh session that has no access to the builder's reasoning. Two axes, standards and spec, run as independent sub-agents. It flags; it never fixes. A fail returns to `propose-fix`. Use after `propose-fix`,…
原文语言：英语
The gate in the self-healing chain. Decides whether a diagnosed failure may proceed to an automated fix, by asking two questions — is this a known pattern in incident memory, and how wide is the blast radius. Known and low-radius may proceed; novel or…
原文语言：英语
Reach a safe state before diagnosing. Halt the affected job, revert to last known good, or stop the loop — containment, never a fix. Records what was contained and what was deliberately left alone. Use immediately after `prove-the-failure`, or any time a…
原文语言：英语
Find root cause by running the red loop from `prove-the-failure` and changing one variable at a time — never by reading the code and reasoning about it. Outputs a written cause with file and line evidence. Use after containment, before any repair is designed.
原文语言：英语
Add the check, test, or workflow guard that stops this failure recurring. May propose a fence change; may never apply one. Use after `verify` passes and its cooldown has elapsed.
原文语言：英语
Write symptom, cause, fix, reviewer verdict and outcome to a searchable store — successes and failures alike. This is the store `classify` reads to decide known versus novel. Use as the final step of every self-healing chain, including chains that stopped…
原文语言：英语
Produce the repair as a diff plus a regression test, never applied directly. Bounded to three attempts, each materially different, then it stops and reports the diagnosis rather than the error. Use only after `classify` returns proceed.
原文语言：英语
Build a tight, red-capable feedback loop that drives the actual failure path and asserts the exact symptom, before any hypothesis is formed. One command, already run once, deterministic, fast, agent-runnable. Use at the start of any self-healing chain, or…
原文语言：英语
Run the loop from `prove-the-failure` and confirm red goes green, then hold a cooldown before any further action is permitted. Use after `adversarial-review` returns PASS, before `immunise`.
原文语言：英语
Use before any spec or non-trivial change becomes code — dispatches the principal-engineer bench (17 specialist seats) to return the engineering requirements the author did not know to ask for, merges them into a machine-checked engineering.md, and gates the…
原文语言：英语
Fail-closed PreToolUse gate on Supabase MCP writes. Read tools stay open; every write is denied unless a founder break-glass file names the project. Fires automatically on any mcp__*Supabase* call — you do not invoke it. Read this when a Supabase write is…
原文语言：英语
Senior Project Manager command (/spm). Use before implementation to turn a rough task, feature, bug, idea, ticket, PR, or repo area into a decision-grade spec.md — via read-only project inspection, a 15+ year specialist board, judge-style challenge,…
原文语言：英语
Use BEFORE any heavy multi-agent, corpus-mining, research or batch task to split work across the whole Nexus fleet instead of one machine. Probes every node for real free capacity and which agent runtimes actually authenticate there, then dispatches jobs to…
原文语言：英语
HeyGen specialist agent — owns avatar / talking-head / spokesperson video via the HyperFrames-by-HeyGen MCP, with a structured status contract the video-director and specialist-council can consult. Use for presenter-led explainers, multilingual avatar…
原文语言：英语
Fable-5 nexus copywriter (15+ yr calibration). The estate's standing content-quality gate — any client- or public-facing words (LinkedIn post, blog/Hub article, email, landing-page/ad copy, video/Remotion script, founder thought-leadership, proposal, deck…
原文语言：英语
Apply the locked-in Unite-Group editorial overlay system to any NotebookLM-generated video (Video Overview output, daily briefings, audio overviews rendered as video). Use AFTER video-use produces a base render; this skill applies a post-process editorial…
原文语言：英语
Production-Readiness Architect command (/readiness-architect). Use to judge whether a project, feature, or release is GENUINELY production-ready — not "renders / CI-green / claimed done". Read-only: inspects the system, grades every capability by evidence,…
原文语言：英语
Use before a multi-step task when it is not obvious which skills apply, or when you catch yourself about to read the 243-skill README to decide. Dispatches a throwaway sub-agent that reads ONLY the router (index.md) and returns the MINIMAL set of skills to…
原文语言：英语
Use before pushing a branch or opening/updating a PR on a GitHub Free PRIVATE repo, and whenever Actions jobs fail to start with "recent account payments have failed or your spending limit needs to be increased" or private-repo CI dies mid-month. Makes a…
原文语言：英语
Deprecated legacy alias retained for audit and rollback. Use pidev-specialised/session-handoff, which owns the canonical continuity capsule and live-state reconciliation contract.
原文语言：英语