Turn a freeform work request into a standardized brief block that team-lead's Pre-flight HARD GATE consumes — collapsing goal verification to a single confirm. Parses the request, derives every brief field it can support, asks ONE batched AskUserQuestion round only for genuinely underdetermined fields, then emits the block verbatim and stops. Standalone operator-intake aid; writes no files, spawns nothing. Trigger: "brief", "create brief", "standardize this request".
Evolve agent definitions in agents/*.md via multi-agent self-review. Phase 0 includes a per-agent historical audit of recent Claude Code transcripts, history, agent memory, and stall signals (TeammateIdle, -r2 respawns, shutdown-rejection). Trigger: "evolve agents", "improve agents", "grow the team", "refine agents".
Audit coherence between agents/*.md and the skill ecosystem (skills/* + .claude/skills/*) across four dimensions, emit a Coherence Report + Remediation Manifest, and ROUTE fixes to evolve-agents/evolve-skills. REPORT-AND-ROUTE ONLY: despite the evolve- prefix it NEVER edits any agent/skill file and NEVER self-invokes the evolve-* skills. Run as a post-edit gate after standalone evolve-agents/evolve-skills edits (evolve-suite runs it automatically). Trigger: "evolve coherence", "audit coherence", "check agent/skill coherence", "cross-reference audit".
Review and improve the Claude Code configuration defined in the repo's Rust builders (src/user.rs, src/user/claude_code.rs) and related scripts via multi-agent self-review. Phase 0 includes a historical audit of recent Claude Code transcripts, history, and agent memory plus a git-history audit of the config sources. Evolves the config SOURCE (Rust builders), NOT live settings.json — for a one-off settings.json/permission/env edit use the bundled update-config skill (/config). Trigger: "evolve config", "improve config", "review config", "refine config sources".
Full team-spawning evolve-* orchestrator that collects LOCAL per-spawn + REMOTE aggregate Claude Code model metrics, categorizes every spawn into a task-tier → model-tier taxonomy, and applies evidence-grounded, operator-gated model-routing edits to team-lead.md. Trigger: "evolve model distribution", "improve model routing", "model distribution", "audit model tiers".
Review and improve skill definitions via parallel @staff-engineer agents. Evaluates 8 dimensions, enforces Content Gate and byte budget. Phase 0 includes a per-skill historical audit of recent Claude Code transcripts, history, and agent memory. Trigger: "evolve skills", "improve skills", "refine skills".
Author a single Architecture Decision Record at docs/adr/{NNNN}-{slug}.md. Loaded into the calling agent's context; the agent drafts the ADR per the format authority below. Trigger: "create ADR", "record this decision", "draft an architecture decision record", "log architectural decision".
Conduct a code review on a scoped artifact (PR, branch, uncommitted, staged, or files). Loaded into the calling agent's context; the calling agent applies the role-appropriate playbook — @staff-engineer runs the 6-dimension general review, @security-engineer runs the security-dimension review. The format authority for both roles' output lives here. NOT the bundled /code-review skill (which can edit the working tree via --fix); this project skill was renamed away from "code-review" to avoid that collision. Emits a structured verdict into the calling agent's context only — it does NOT post to the PR; to post findings as inline PR comments in the operator's voice use Skill(review-and-comment). Trigger: "code review", "review this PR", "review the diff", "security review of changes".