Recursive Skill Failure-Fix Loop
Every time your skill fails on a real task, feed the failure back and have Claude update the skill.md so it never breaks that way again.
Before & after
Same prompt, same agent — the only difference is the skill.
That sounds frustrating. You could try a few things—go back and review what the skill was supposed to do, check if there are edge cases you missed, maybe simplify the instructions. You might also want to test it with different inputs to see where it fails. It could help to document what's breaking so you can update the skill gradually. Would you like to walk through the failure together and see what's going wrong?
RECURSIVE SKILL FAILURE-FIX LOOP STEP 1: CAPTURE THE FAILURE • Run the skill on a real task (not a test case) • Screenshot or copy the exact error output • Note the input context: what conditions triggered the failure? EXAMPLE FAILURE CAPTURED: Skill: "Pull YouTube analytics into Notion" Error: "API returned 403 Forbidden - channel not verified" Context: First run on a new creator account with YouTube Premium STEP 2: DIAGNOSE ROOT CAUSE Analyze what the skill assumed (e.g., "channel is already verified") that didn't hold. STEP 3: UPDATE SKILL.MD Add a new section to SKILL.md: GATEWAY CHECK: Before pulling analytics, verify the YouTube channel is verified. If unverified, ask the user to complete verification first. Link to YouTube Studio. STEP 4: RE-RUN THE SAME TASK Confirm the updated skill now handles the previously failing case. STEP 5: REPEAT Wait for the next failure (real usage). Feed it back. Update skill.md again. After 5 recursive loops, the skill becomes production-grade. THE PATTERN: Manual walkthrough → Initial skill → Failure 1 → Update → Failure 2 → Update → ... → Failure 5 → Publish Each failure improves the skill's robustness. Never build blindly; learn from each real-world break.
About this skill
name: recursive-skill-failure-fix-loop description: Every time your skill fails on a real task, feed the failure back and have Claude update the skill.md so it never breaks that way again.
Recursive Skill Failure-Fix Loop
What This Does
The creator argues you should never write skills by hand or install blindly. Instead you should manually walk through the workflow, produce a successful run, then tell Claude to create the skill from that context. When the skill later fails, ask Claude for the exact error, feed it back, and instruct it to update the skill.md so the failure is impossible next time.
What You Get
- Manual first-run workflow before skill creation
- failure-capture prompt (what error did you get?)
- skill.md self-update after each failure
- five-loop refinement pattern
- deliberately avoids downloading random skills
Customize
- workflow_first: force a successful manual run before skill creation
- failure_capture: how to prompt for the raw error
- update_scope: only skill.md or references too
- loop_target: number of iterations before publishing
- commit_each_fix: git commit per improvement
Sample Output
The creator's YouTube-report skill pulls Notion, Dub, YouTube Analytics, Twitter, and eight other sources flawlessly after five recursive loops of failure-then-fix.
Target Audience
Skill authors who care about production reliability over shipping speed
SKILL.md preview
---
name: recursive-skill-failure-fix-loop
description: Use this skill to feed a skill's real-world failures back into Claude so it rewrites the skill.md and closes off that failure mode for good.
version: 1.0.0
category: Agent Orchestration
author: AgentVolt
license: proprietary
tags:
- agent-orchestration
- standard
---
# Recursive Skill Failure-Fix Loop
Improves a skill file by running it against a real task, capturing the exact failure when it breaks, and rewriting the skill.md so that failure cannot recur.
## When to use
… (sign up to view the full skill)More ai & agents skills
View all AI & Agents skills →Self-Improving Skill Overnight Loop
Run an automated overnight research loop on a Claude skill: try a tweak, run binary assertions, keep or revert based on pass rate, and iterate.
Ralph Loop Autonomous Overnight Runner
Wire the Claude Code CLI mode (-p flag) into the Ralph loop plugin so a while-loop keeps feeding the same prompt with updated files until the task is done.
Critique Sub-Agent Reviewer
Every time the main agent produces copy, research, or code, hand it to a critique sub-agent that returns fix notes before the output ships.
AI Starter Toolkit
A free, curated starting point — the essential prompts and habits to get real value from AI on day one.