ralph-wiggum-aot

Plugin ralph-wiggum-aot in yodakeisuke/ralph-wiggum-aot. Commands read from the repo's manifest — 1 month ago.

Skills (4)

• aot-dag

  This skill should be used when you need to manipulate the AoT DAG (Work Graph) structure. Use it when decomposing Atoms into sub-tasks, resolving Atoms, managing AND/OR dependencies, performing backtracking on OR branches, or shrinking the DAG by replacing resolved Atoms with summaries in Bindings.

• convergence

  This skill should be used when evaluating loop progress, detecting stalls, deciding whether to continue or stop iteration, or switching strategies when the DAG is not shrinking. Use it in the coordinator agent to determine next actions based on convergence metrics.

• parallel-exec

  This skill should be used when deciding whether to execute multiple Atoms in parallel, detecting side-effect conflicts between concurrent operations, or integrating results from parallel Worker agents. Use it in the coordinator agent when multiple Atoms are ready for execution.

• state-contract

  This skill should be used when you need to read, write, or manipulate the AoT Loop state file (.claude/aot-loop-state.md). Use it when implementing commands like /align-goal, /enter-recursion, /exit-recursion, /redirect, or when agents need to update Atom status, Bindings, or control flags.

Slash commands (5)

• /ralph-wiggum-aot:align-goal

  Goal alignment - Interactive agreement formation for AoT Loop (background, deliverables, completion criteria)

• /ralph-wiggum-aot:enter-recursion

  Start AoT Loop - Begin autonomous iteration after goal alignment

• /ralph-wiggum-aot:exit-recursion

  Stop AoT Loop - Manually stop the autonomous iteration loop

• /ralph-wiggum-aot:help

  Explain AoT Loop plugin commands and usage

• /ralph-wiggum-aot:redirect

  Redirect AoT Loop - Interrupt and modify direction without stopping

Subagents (4)

• aot-coordinator

  AoT Loop coordinator agent that analyzes the Work Graph, decides which sub-agents to spawn (Probe, Worker, Verifier), manages parallel execution, integrates results, and evaluates convergence.

• aot-probe

  AoT Loop probe agent for investigating feasibility of an approach without making changes. Read-only exploration to inform coordinator decisions.

• aot-verifier

  AoT Loop verifier agent that checks if the base_case completion criteria is satisfied. Supports checklist format, negative conditions, and LLM as a Judge for quality evaluation.

• aot-worker

  AoT Loop worker agent that executes a single Atom task. Has full tool access to make changes, run tests, and complete the assigned work.

Hooks (2)

• SubagentStop

  ${CLAUDE_PLUGIN_ROOT}/hooks/subagent-stop-hook.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/permission-request-hook.sh