robin

Plugin robin from marketplace robin in waynewangyuxuan/Robin. Commands read from the repo's manifest — 1 month ago.

Skills (12)

• robin-committer

  AI-Robin Committer. Executes a git commit on behalf of the kernel using the verbatim message provided by Merger or Degrader. Never composes commit messages itself.

• robin-degrader

  AI-Robin Degrader. Writes the context-degraded-*.yaml spec narrative and updates escalation-notice when a scope is degraded. Reads original specs + ledger history to compose the narrative (domain work the kernel cannot do itself).

• robin-executor

  AI-Robin Executor stage. Given a single task (typically one milestone), loads relevant context, writes/modifies code and specs, and returns a structured artifacts summary. Does NOT git commit (kernel delegates that to Committer).

• robin-finalizer

  AI-Robin Finalizer. Generates the end-of-run delivery bundle (.ai-robin/DELIVERY.md) summarizing what was built vs degraded, by scanning ledger + reading intent specs.

• robin-intake

  AI-Robin Intake stage. Conducts the one-time user Q&A at the start of a Robin run — surfaces decisions, fills gaps, produces planning-ready Feature Room specs. Only human-facing stage of an AI-Robin run.

• robin-kernel

  AI-Robin kernel (main agent). Autonomous multi-agent workflow that takes a one-shot human intake and delivers a software project end-to-end. Dispatches sub-agents (intake, planner, scheduler, executor, researcher, reviewer, merger, committer, degrader, finalizer) and routes their return signals. Loa

• robin-merger

  AI-Robin Merger. Consolidates N review sub-verdicts into one merged verdict with a composed git commit message for the Committer.

• robin-planner

  AI-Robin Planning stage. Turns Intake's specs into an executable plan with milestones, module boundaries, and API contracts. May re-spawn for research gaps, sub-planning, or post-review rework.

• robin-researcher

  AI-Robin Researcher. Answers a specific factual question Planner needs to finalize a plan. Uses web search + targeted analysis to produce structured findings with confidence.

• robin-review-planner

  AI-Robin Review-Planner. Given a batch's change artifacts, decides which domain-specific reviewer playbooks to run (code-quality always, plus domain-specific like frontend-a11y, backend-api, db-schema as applicable).

• robin-reviewer

  AI-Robin generic reviewer skill. The shared review flow used by all domain-specific reviewer agents (robin-reviewer-code-quality, robin-reviewer-frontend-a11y, etc.). Defines how to load scope, walk a domain checklist, and emit a structured verdict. Domain-specific checklists live in `domains/{domai

• robin-scheduler

  AI-Robin Scheduler stage. Reads the plan + current progress and decides which milestones to tackle in the next batch, at what concurrency (parallel / sequential / mixed).

Slash commands (3)

• /robin:robin-resume

  Resume an interrupted AI-Robin run, or resolve a paused milestone checkpoint

• /robin:robin-start

  Start a new AI-Robin run on the current working directory

• /robin:robin-status

  Report status of the current AI-Robin run without changing state

Subagents (11)

• robin-committer

  AI-Robin Committer Agent. Executes a git commit using the exact message provided by the trigger signal. Invoked only by the AI-Robin kernel after review_merged or degradation_spec_written.

• robin-degrader

  AI-Robin Degrader Agent. Writes the context-degraded-*.yaml spec with narrative and updates escalation-notice.md. Invoked only by the AI-Robin kernel when a scope is degraded.

• robin-executor

  AI-Robin Executor Agent. Writes actual application code for one milestone. May modify source files and run tests. Does NOT git commit (kernel does that). Invoked only by the AI-Robin kernel.

• robin-finalizer

  AI-Robin Finalizer Agent. Generates the end-of-run delivery bundle (DELIVERY.md). Invoked only by the AI-Robin kernel on all_complete.

• robin-intake

  AI-Robin Intake Agent. Intake stage — the only sub-agent that interacts with the user. Do not invoke for general intake tasks; only invoke as part of an AI-Robin dispatch loop.

• robin-merger

  AI-Robin Merger Agent. Consolidates N review sub-verdicts into one merged verdict with a composed git commit message. Invoked only by the AI-Robin kernel after all review playbooks return.

• robin-planner

  AI-Robin Planner Agent. Turns Intake's intake into an executable plan with milestones, modules, and contract specs. Invoked only by the AI-Robin kernel.

• robin-researcher

  AI-Robin Researcher Agent. Answers a specific factual question via web search. Invoked only by the AI-Robin kernel when Planning returns planning_needs_research.

• robin-review-planner

  AI-Robin Review Planner Agent. Meta-agent that decides which review playbooks to run for a batch. Invoked only by the AI-Robin kernel at the start of every review stage.

• robin-reviewer-code-quality

  AI-Robin code-quality reviewer. Always-on review that evaluates correctness, readability, maintainability, error handling, testing, and basic security across changed code. Uses the robin-reviewer generic flow with code-quality domain rules.

• robin-scheduler

  AI-Robin Scheduler Agent. Scheduler — decides which milestones run in the next batch and at what concurrency. Read-mostly. Invoked only by the AI-Robin kernel.

Hooks (5)

• PreToolUse

  python3 $CLAUDE_PLUGIN_ROOT/hooks/pre_task.py

• PostToolUse

  python3 $CLAUDE_PLUGIN_ROOT/hooks/post_task.py

• SessionStart

  python3 $CLAUDE_PLUGIN_ROOT/hooks/session_start.py

• Stop

  python3 $CLAUDE_PLUGIN_ROOT/hooks/stop.py

• SubagentStop

  python3 $CLAUDE_PLUGIN_ROOT/hooks/subagent_stop.py