grade-iterate

Editorial and source context

What this Skill does

Grade Iterate is a source-indexed Agent Skill for design & frontend work. Based on the SKILL.md description, it focuses on phase 3 of building a claude managed agent — the bounded grade→iterate loop. Use the linked GitHub file to confirm scope and prerequisites before enabling it.

What the author says

Phase 3 of building a Claude Managed Agent — the bounded grade→iterate loop. Define a CMA outcome (a required markdown rubric graded by an isolated grader), read each verdict, decide the next move (sharpen / re-run / promote to schedule), and once a version passes, run held-back eval cases in parallel. Use when the user says "grade my agent", "make it pass the rubric", "iterate until it's good", "is it good enough", or when the orchestrator routes phase=grade-iterate. outcome_builder.py builds the user.define_outcome payload (rubric required, max_iterations clamped 1..20 — never unbounded); verdict_reader.py reads the grader result and recommends the next move; eval_scaffold.py generates held-back cases + a parallel run plan (capped at the 25-thread CMA ceiling). Distinct from stage-launch (first launch) and run-without-you (scheduling).

Compatibility and requirements

• Use an agent runtime that supports Agent Skills and the linked SKILL.md format.
• Confirm the source repository's tools, SDKs, and platform prerequisites for ui and frontend work.

How to install or import it

• Open the linked GitHub SKILL.md and review its scope and setup instructions.
• Install or import the skill using the agent runtime's documented workflow.
• Run a small, non-sensitive test and verify the output before broader use.

Permissions and risks

• Review every command, file path, network request, dependency, and credential scope before enabling it.
• GitHub stars indicate popularity, not safety or correctness; validate the source and test with non-sensitive data.

Example workflows

• Ask an AI agent to apply Grade Iterate to ui and frontend work described in the source record.
• Have the agent state assumptions, required tools, and expected output before using this design & frontend skill.
• Compare the result with the linked GitHub SKILL.md and verify it against your project requirements.