菜单
SkillsMP 已收集 aiosbrain/aios-workspace 中的 64 个 Skill。打开任一 Skill 可查看来源和详情。
已展示 40 / 64 个已收集 Skill。
Manage the AIOS Linear board (the ONLY PM tool — Plane is retired). Use whenever updating, reading, or commenting on AIOS Linear issues (identifiers like AIO-72, AIO-75), the brain→PM projection-tracking epic, or the backlog. Triggers on "update AIO-NN", "the…
原文语言：英语
Use YOUR own Slack to act as you — send messages and DMs to teammates, read your channels/DMs, and react — via the `slack` CLI (a zero-dependency tool that talks straight to the Slack Web API with your USER token). Messages post AS you and replies land in…
原文语言：英语
Verify that a fix or change actually did what it was for, before reporting that it works. Use before any "it works", "verified", "fixed", "confirmed", or "done" about something that was previously broken — and whenever a hook or a reviewer asks you to…
原文语言：英语
Analyze recent Codex session history, actual skill-loading evidence, resolver usage, installed skill catalogs, and accumulated Claude instincts to identify repeated friction, missed skill-routing opportunities, and reusable workflow improvements. Use when…
原文语言：英语
Pull Granola meeting notes and transcripts into 1-inbox/transcripts/ by calling the Granola API directly (this is our own connector — it does NOT use the Granola MCP). Use when the user says "pull my Granola notes", "get today's meeting transcripts", "ingest…
原文语言：英语
Propose 3–5 non-overlapping agent workstreams from the AIO Linear board. Use before starting parallel agent batches, when planning a factory work session, or when the user asks for a workstream update / batch plan. Triggers on "workstream update", "batch…
原文语言：英语
Audit whether a candidate spec is ready against an immutable repository baseline and the AIOs readiness rubric. Use for “is this spec ready,” readiness evaluation, or repository-backed spec audits; never revise or publish the candidate.
原文语言：英语
Build a presentation-grade, brand-themeable HTML slide deck — demo, proposal, update, or pitch — from a pre-debugged component library instead of a blank page. Ships three themes, a slide catalog with copy-pasteable markup, a brand intake schema for adding…
原文语言：英语
Query and update Linear issues, projects, and cycles via the Linear GraphQL API using a personal API key (our own connector — Linear's MCP is OAuth-only). Use when the user asks about their Linear issues, sprint/cycle status, or wants to create or update an…
原文语言：英语
Use ast-grep (sg) for AST-aware code search and deterministic rewrites/codemods across many languages. Trigger when the target is syntax SHAPE, not text: find every function/call/class/import shaped like X, rewrite console.log to logger.info, strip `as any`,…
原文语言：英语
Classify every unmerged remote branch as truly-unmerged, already-shipped-under-a-different-hash (squash-merge duplicate), or genuinely stale — with evidence, not guesswork. Use when asked "are these branches actually stale", "check unmerged branches", "clean…
原文语言：英语
Fresh-context review of a diff with severity-triaged findings (P1/P2/P3). Use when the user asks to "review" a change/PR/diff, or as the standard gate on any agent-authored change before merge. The reviewer must NOT be the session that wrote the code.
原文语言：英语
The closing step of every task — codify what was learned so the system gets better each use. Use after completing any task where the agent was corrected, a bug was found, a procedure was discovered, or friction repeated. Also on request ("capture this", "make…
原文语言：英语
Use whenever a task needs a commit or a git-history investigation. Covers atomic commits, staging by hunk, detecting commit-message style, rebase/squash/fixup/autosquash, blame, bisect, reflog, git log -S/-G, and questions like who wrote this or when was this…
原文语言：英语
Research → Plan → Implement discipline for any non-trivial change. Use BEFORE writing code when a task touches more than one file, introduces a dependency, changes behavior users see, or when the user asks to "plan", "spec", or "design" something. Skip only…
原文语言：英语
Coding-standards conscience for writing or editing code in a typed language (Python, TypeScript, Rust, Go, …). Use when authoring non-trivial code and you want strict types, honest module size, and tests that assert behavior not prose. A code-smell +…
原文语言：英语
Structured restructuring of EXISTING code — extract, split, rename, modernize, de-duplicate, move responsibilities — with zero behavior change and a test run after every step. Use when the task is to reshape working code (not add a feature, not the post-diff…
原文语言：英语
Behavior-preserving cleanup pass over a reviewed diff — run AFTER tests pass and review is addressed, BEFORE merge. Deletes dead code, collapses needless indirection, applies YAGNI. Use when the user says "simplify", "clean up this diff", or as the standard…
原文语言：英语
Write or improve a SKILL.md so it actually triggers and actually changes agent behavior. Use when the user says "create a skill", "turn this into a skill", "why isn't this skill triggering", or when compound-learnings decides a lesson deserves a skill.
原文语言：英语
Four-phase root-cause debugging loop for anything broken, throwing, failing, flaky, or slow. Use whenever the user reports a bug or a test fails unexpectedly — BEFORE proposing any fix. Root cause first; a fix without a diagnosis is a guess.
原文语言：英语
Enforced RED-GREEN-REFACTOR test-driven development. Use when implementing any behavior change that can be expressed as a test — new features, bug fixes, edge cases. The test MUST fail first; a test that passes on first run proves nothing.
原文语言：英语
Find test files that exist on disk but run in neither the local test command nor CI (orphaned tests = false confidence), plus stale coverage reports. Use when asked "is this test actually running", "check test wiring", "orphaned tests". Flags and suggests…
原文语言：英语
Verify a change actually works by exercising it end-to-end and observing real behavior — not just green tests. Use before declaring any nontrivial change done, before opening a PR, or when the user asks "does this actually work?".
原文语言：英语
Rigorous visual QA for any UI you built or changed — web pages, components, and terminal/TUI output. Use AFTER a UI change and BEFORE calling it done, or when asked does this look right / match the mock / is the layout broken / does the CJK text clip.…
原文语言：英语
Place an engineer on the Agentic Maturity model (5-level spine × 5 axes) via a short interview, apply the verification cap, and prescribe the 2-3 highest-leverage patterns to practice next. Use when someone asks "assess my agentic maturity", "what should I…
原文语言：英语
Context-engineering hygiene audit — contract bloat, context-utilization discipline, skill misfires, MCP tool sprawl. Use on a weekly/monthly cadence, when sessions feel slow or agents "get dumber over time", or before onboarding new engineers onto the harness.
原文语言：英语
Token/cost visibility for agentic engineering — per session, per day, per model lane. Use when someone asks about token usage, AI spend, "which lane is the money going to", or as a weekly habit once multi-model routing is live. Read-only reporting; changes…
原文语言：英语
Draft or restructure a Linear issue, implementation spec, acceptance criteria, or issue slice into a cold-start buildable increment. Use for “write the spec,” “make this issue buildable,” or equivalent authoring requests; do not use to evaluate, repair,…
原文语言：英语
Change install, onboarding, workspace hydration, config merge, skill installation, or uninstall behavior. Use to preserve user-authored state while testing idempotency, collisions, symlink or path escape, modes, dependencies, partial failure, rollback, and…
原文语言：英语
Change `aios update`, toolkit pull, re-exec, contribution worktrees, scheduler refresh, or version handoff behavior. Use to preserve preview and mutation boundaries, dirty state, source proof, runtime environment, recovery, and fail-closed semantics.
原文语言：英语
Close or abandon a delivery worktree after merge or blockage. Verify remote state, ancestry, post-merge behavior, and protected primary state before classifying cleanup as safe, blocked, or owner-decision; never delete dirty or unproven work.
原文语言：英语
Change a versioned API, event, transcript, persisted record, schema, or wire contract. Use when compatibility, migrations, producers, consumers, validators, fixtures, or rollback and read behavior must be designed and sequenced.
原文语言：英语
Publish or replace an AIO Linear issue description from a locally evaluated candidate. Use only when the user explicitly asks to publish and provides or approves the exact issue and candidate; never auto-trigger from review, readiness, authoring, or repair…
原文语言：英语
Change transcript ingestion, extraction, redaction, provenance, decision or task derivation, or audience-tier output. Use to enforce source allowlisting, grounding, redaction-before-egress, tier boundaries, containment, replay safety, and extraction…
原文语言：英语
Repair a candidate spec only after concrete readiness findings exist or the user explicitly requests repair. Convert findings into bounded edits while preserving fixed decisions, capability truth, scope, and intent; do not publish or claim readiness without…
原文语言：英语
Begin implementation safely in an AIOs repository when a task names or needs a branch or worktree. Resolve the repository, issue, owner, base, and branch; snapshot primary dirty state; and hand off a hydrated worktree without stashing, resetting, cleaning, or…
原文语言：英语
Reconcile Local Bugbot, CodeRabbit, CI, GPT review, security review, and plan-conformance findings for one exact PR head. Use after deterministic findings collection when evidence is ambiguous; do not re-review code, fix, merge, or treat a green check as…
原文语言：英语
Verify source changes that require rebuilt `dist/`, vendored assets, generated catalogs, Tauri resources, manifests, or lock artifacts. Use to identify source-to-generated relationships, rebuild canonically, prove freshness and parity, and test packaged…
原文语言：英语
Verify claims about lifecycle hooks, prompt injection, tool interception, stop continuation, subagents, or cross-runtime parity. Use for runtime or adapter preflights and capability matrices before a spec relies on behavior that may differ by installed…
原文语言：英语
Prove parity when a JSON Schema, TypeScript or runtime validator, generated type, or fixture set changes. Use to detect optionality, enum, default, coercion, unknown-field, and test-wiring drift; do not decide product contract semantics.
原文语言：英语