magi

Plugin magi from marketplace cc-magi-engineer in MagdielCAS/cc-magi-engineer. Commands read from the repo's manifest — 1 month ago.

Skills (1)

• agentic-loop

  Run a bounded, verifiable agentic loop that plans, implements, and evaluates work until it passes or the step budget is exhausted. Use when the user asks for an "agentic loop", "autonomous loop", "self-correcting loop", "plan/implement/evaluate loop", "run until it passes", "keep iterating until X",

Subagents (3)

• agentic-loop-evaluator

  Evaluator (LLM-as-judge) role for the agentic-loop skill. Independently verifies an Implementer's output against a fixed success criterion and returns a BINARY verdict — PASS or FAIL — plus a detailed, evidence-cited critique. Never emits numeric scores or middle tiers. Uses the user's calibration e

• agentic-loop-implementer

  Implementer role for the agentic-loop skill. Executes exactly ONE subtask — writes code, edits files, runs commands — writes its full diff/trace to the given result file, and returns a compact pointer plus a self-check. On a retry it receives the previous Evaluator FAIL critique as a verbatim work o

• agentic-loop-planner

  Planner role for the agentic-loop skill. Decomposes a high-level goal into an ordered list of small, independently verifiable subtasks, each mapped to a concrete success criterion. Also re-decomposes the remaining work when a subtask has failed repeatedly (escalation). Read-only: it plans, it does n

Similar plugins