Clade

Clade goal-driven autonomous improvement loop (Blueprint architecture — deterministic pre/post phases + LLM supervisor/worker nodes, converges when goal met or max-iter hit). NOT the Claude Code built-in /loop (which polls a prompt on an interval like `/loop 5m /foo`) — if the user wants interval polling, route to the built-in.

Adaptive, resumable Git delivery across repositories and agent runtimes — probe policy, checkpoint coherent work, publish, review, integrate, and verify cleanup

Adaptive, resumable Git delivery across repositories and agent runtimes — probe policy, checkpoint coherent work, publish, review, integrate, and verify cleanup

Inspect or select a Clade execution connection while keeping agent runtime, inference provider, wire protocol, and model distinct. Use for Claude/Codex runtime selection, Anthropic/OpenAI/MiniMax/Kimi/custom gateways, model routing, or connection troubleshooting.

Inspect or select a Clade execution connection while keeping agent runtime, inference provider, wire protocol, and model distinct. Use for Claude/Codex runtime selection, Anthropic/OpenAI/MiniMax/Kimi/custom gateways, model routing, or connection troubleshooting.

Equipment manager for Claude Code projects — inventory local skills/agents/scripts, audit external upstream repos (red-flag + prompt-injection screening), and sync selectively after user approval. Absorbs container-layout skill repos AND single-skill-at-root repos (e.g. a company design-system repo). Project-agnostic.

Integrate an exact reviewed PR under repository policy, choose truthful history semantics, and verify cleanup

Integrate an exact reviewed PR under repository policy, choose truthful history semantics, and verify cleanup

Show Codex rate-limit usage and pace, or configure the native Codex status line

Create repository-adaptive checkpoint commits for coherent work, including requests to split changes into commits and push; publication remains separately authorized

Publish or update one repository-adaptive, exact-SHA pull request without duplicating or broadening authority

Security audit skill — systematic OWASP + STRIDE review of a project. Covers attack surface, secrets archaeology, dependency supply chain, OWASP Top 10, threat modeling, and false-positive filtering. Outputs a prioritized findings report.

Post-ship documentation sync — updates README, CHANGELOG, AGENTS.md, ARCHITECTURE, and TODOS after a release. Ensures docs never drift from code. Run after /commit or /ship before closing a PR.

Create distinctive, production-grade frontend interfaces with high design quality. Detects and enforces the project's design system (.design-system.md, design-system skill repo, or DESIGN.md) — hard-rule grep checks, review checklist, decisions log; can also author a new design system (SKILL.md + DESIGN.md + assets).

Accept the recommendation from the most recent A/B/C option set and execute it without re-asking. Use when the user says "按你推荐的来" / "go with your pick" — they glanced at the options, the recommendation looked right, they don't want to re-read all of them.

End-of-session context handoff. Saves session state so the next session or a parallel agent can pick up exactly where the left off. Run this when context is getting full (~80%) or before stopping work.

Incident response mode — diagnose a production issue, write a postmortem, add follow-up tasks to TODO.md. Use when user says "/incident [description]".

Root cause analysis skill — diagnose bugs systematically before fixing. Iron Law: no fix without a confirmed hypothesis. Uses scope lock, 3-strike escalation, and structured debug reports.

Scan project structure and generate ARCHITECTURE.md with a Mermaid module relationship diagram. Use when user says "/map".

\"What should I work on next?\" — two modes. Fast mode (default): one-shot recommendation from docs + git state, ≤10 lines, no interview. Deep mode (`/next deep` or when user asks to explore priorities): multi-round interview to surface the best move.

Project orchestrator mode — ask clarifying questions, decompose goals into tasks, write proposed-tasks.md for workers to execute. Use inside the Orchestrator Web UI.

Start-of-session context pickup. Reads the latest handoff file and resumes work where the last session left off. Run this at the start of a new session if a handoff file exists.

Research external tools/competitors for a topic and write findings to BRAINSTORM.md (project topics) or ~/.clade/research/ (personal topics). Use when user says "/research [topic]".

Engineering retrospective — generates a data-driven report of development velocity, commit patterns, hotspots, and team insights from git history. Saves results for week-over-week comparison.

Review an exact PR head in isolation using trusted-base instructions and execution evidence

Clade coverage-driven project review — walks every VERIFY.md checkpoint, fixes failures in-session, converges when all checkpoints pass. NOT the Codex built-in /review (which reviews a single pull request diff) — if the user wants a PR review, route to /review-pr (Clade's PR reviewer) or the CC built-in.

Show a provider-neutral, freshness-aware snapshot of active Agent work, Git delivery, execution identity, and usage limits. Use for “what's going on right now”, what is running, progress updates, active runtime/provider/model, or stuck-work checks.

End-of-session documentation sync — updates TODO.md and PROGRESS.md only (run /commit after to commit everything)

Verify project behavior anchors — compilation, tests, and interaction checks after autonomous runs. NOT the Codex built-in /verify (which runs the app to observe a single change working) — this one walks the AGENTS.md "Features (Behavior Anchors)" list.

Create or inspect runtime-adaptive isolated Git workspaces with explicit ownership and delivery routing

Show a provider-neutral, freshness-aware snapshot of active Agent work, Git delivery, execution identity, and usage limits. Use for “what's going on right now”, what is running, progress updates, active runtime/provider/model, or stuck-work checks.

Create repository-adaptive checkpoint commits for coherent work, including requests to split changes into commits and push; publication remains separately authorized

Publish or update one repository-adaptive, exact-SHA pull request without duplicating or broadening authority

Review an exact PR head in isolation using trusted-base instructions and execution evidence

Create or inspect runtime-adaptive isolated Git workspaces with explicit ownership and delivery routing

Project orchestrator mode — ask clarifying questions, decompose goals into tasks, write proposed-tasks.md for workers to execute. Use inside the Orchestrator Web UI.

Full release pipeline — tests → coverage → review gate → version bump → CHANGELOG → commit → PR

Orchestrate a fleet of parallel `codex exec` workers with you (Claude Code) as the supervisor — spawn one per isolated git worktree, dispatch headless, verify each INDEPENDENTLY, PR/merge. The manual "codex-ultracode" pattern for fanning out real implementation, research, or review work onto Codex. Bakes in the hard gotchas (stdin blocking, background tracking, don't-trust-self-reports, writer isolation). Triggers on — orchestrate codex, codex workers, codex fleet, spawn codex, delegate to codex in parallel, manual ultracode, 开 codex 小弟, 派 codex worker — NOT for a single cross-vendor opinion (use the `second-opinion-codex` agent), NOT for web-UI worker decomposition (use `/orchestrate`).

Show Codex rate-limit usage and pace, or configure the native Codex status line

Scaffold a new Clade skill end-to-end — interviews for use cases and trigger phrases, generates SKILL.md + prompt.md with spec-validated frontmatter (bilingual triggers, NOT-for disambiguation), wires golden-set routing tests, and runs the lint gate before committing