Loading...
Loading...
Found 17 Skills
Compress agent-facing instructions to the fewest words that preserve behavior, constraints, and clarity.
Structured session analysis and project instruction refinement using a five-type intervention taxonomy (Correction, Repetition, Role Redirect, Frustration Escalation, Workaround) with severity scoring to categorize process gaps. Refines project instructions (CLAUDE.md, AGENTS.md, .team/coordinator-instructions.md) with structural (not advisory) language, maintains WORKING_STATE.md for crash recovery (read-first-after-any- interruption protocol), and implements a self-reminder protocol (re-read constraints every 5-10 messages to prevent role drift). Includes advisory- to-structural promotion pattern for recurring gaps. Activate after milestones, repeated user corrections, session restarts, crash recovery, every 5 completed tasks, or on user request. Triggers on: "reflect on this session", "why do I keep correcting you", "update project instructions", "update working state", "session retrospective", "crash recovery", "context compaction", "role drift", "I keep telling you the same thing", "analyze my corrections". Also relevant when the agent notices repeated corrections, needs to resume after compaction, or wants to prevent known failure modes from recurring.
Autonomously optimize an existing AI skill by running it repeatedly against binary evals, mutating one instruction at a time, and keeping only changes that improve pass rate. Based on Karpathy-style autoresearch, but applied to SKILL.md iteration instead of ML training. Use when optimizing a skill, benchmarking prompt quality, building evals for a skill, or running self-improvement loops on reusable agent instructions. Triggers on: skill-autoresearch, optimize this skill, improve this skill, benchmark this skill, eval my skill, run autoresearch on this skill, self-improve skill.
A methodology for iteratively improving agent-facing text instructions (skills / slash commands / task prompts / CLAUDE.md sections / code-generation prompts) by having a bias-free executor actually run them and evaluating two-sidedly (executor self-report + instruction-side metrics). Keep iterating until improvements plateau. Use it right after creating or substantially revising a prompt or skill, or when you want to attribute an agent's unexpected behavior to ambiguity on the instruction side.
Use when agent instruction files (AGENTS.md, rules/) need analysis, trimming, or restructuring. Orchestrates /imperatives → /policy-algebra → /visualize into a distillation pipeline.