factory

Plugin factory from marketplace onur in onurcelep/ai-factory. Commands read from the repo's manifest — 2 days ago.

Skills (7)

• ci-agent-ops

  Operating and diagnosing the @claude CI pipeline. Use when a @claude workflow run or auto PR review misbehaves (green check but no PR opened, no review note, instant finish, silent failure), when verifying CI-agent health, or before changing claude*.yml workflows or the CLAUDE_CODE_OAUTH_TOKEN secre

• factory-init

  Stamp the current repository with the standard AI-dev setup - @claude workflows, .claude/settings.json plugin wiring, marker-fenced CLAUDE.md, AGENTS.md. Use when initializing a new repo or onboarding an existing repo onto the ai-factory standard.

• factory-status

  Fleet check - list every repo of the current GitHub user that is factory-stamped and report whether each is current or stale against the marketplace's latest version. Use when asking "which of my repos need a factory-update", after templates change, or to verify propagation worked.

• factory-update

  Refresh the ai-factory standard parts of an already-initialized repo - the @claude workflows, plugin wiring in .claude/settings.json, and the marker-fenced standard block in CLAUDE.md. Use after the ai-factory templates change. Never touches project-specific content.

• model-routing

  Model routing policy for token efficiency. Use when spawning subagents, choosing which model tier (haiku, sonnet, opus) an implementer, reviewer, or research agent should run on, or editing model pins in @claude GitHub workflow files.

• release-flow

  Standard branch-and-PR release discipline (trunk-based, PR-per-change) for local work and the remote @claude CI agent. Use before branching, opening a PR, merging, deploying, or when acting as the @claude GitHub Actions agent.

• repo-memory

  Repo-committed agent memory (docs/memory/) shared by local sessions, Claude Code cloud, and the @claude GitHub Action. Use when reading or recording durable project learnings, promoting local auto-memory into the repo, or deciding where a learned fact belongs.

Subagents (3)

• factory-implementer

  Cheap implementer for fully-specified tasks — the plan or brief already contains the complete change (code, exact edits, or unambiguous steps) and the job is transcription plus running the stated checks. Use PROACTIVELY when executing plan tasks that need no design judgment. Not for debugging, desig

• factory-researcher

  Deep researcher for questions where answer quality dominates cost — architecture and design investigation, upstream/API capability research, root-cause analysis across a codebase, alternatives evaluation. Use when the task is research or design, not implementation. Expensive by design; do not use fo

• factory-reviewer

  Mid-tier reviewer and fixer — reviewing a completed task against its brief, applying review findings, judgment calls on scoped changes. Use PROACTIVELY to review a subagent's work before integrating it, or to fix findings from a review. Not for whole-branch/final review (use the strongest model in t

Hooks (1)

• PreToolUse

  bash "${CLAUDE_PLUGIN_ROOT}/hooks/protect-main.sh"