skills.rest
chrisschouk
chrisschoukCOMMUNITY
Explore all skills by this author →
Website𝕏@chrisschoukGitHub

eval-loop

Documents task failures in a corrections log so future agents avoid repeating mistakes.

Updated Jul 29, 2026
Search Tags:#root cause#continuous improvement#corrections#agent evaluation#marketing memory#failure logging
One-click install
npx skills add https://github.com/chrisschouk/chrisscho-uk-marketing-skills --skill eval-loop-chrisschouk
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: eval-loop
Source: https://github.com/chrisschouk/chrisscho-uk-marketing-skills/tree/main/skills/shared/eval-loop
Command: npx skills add https://github.com/chrisschouk/chrisscho-uk-marketing-skills --skill eval-loop-chrisschouk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When an AI agent fails a task (bounced outreach, hallucinated data, wrong brand voice), the mistake often gets repeated by the next agent because nothing was recorded. This Skill captures each failure in a structured corrections log so the same error never happens twice. ## Core Features & Use Cases - Structured Failure Logging: Records failures in growth-os/eval/corrections.md using a three-part format: what went wrong, root cause, and fix. - Actionable Root Cause Analysis: Enforces specific, actionable root causes and concrete fixes rather than vague explanations. - Use Case: After a cold email bounces because the agent missed an exclusion list, log the failure with the exact cause (didn't read excludes.md) and the fix (always check excludes.md before outreach), so the next agent checks the file first. ## Quick Start Document the failed outreach task in growth-os/eval/corrections.md using the three-part corrections format.

Frequently Asked Questions about eval-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I document AI agent failures so they don't repeat?▼

Log each failure in growth-os/eval/corrections.md using a three-part format: what went wrong, the root cause, and the fix. Keep entries as terse bullet points with specific outcomes and concrete corrective steps.

What is a corrections log for AI agents?▼

A corrections log is a markdown file recording task failures with their root causes and fixes. Future agents read it before acting, so documented mistakes like wrong pricing or missed exclusion lists are not repeated.

When should I write a correction entry?▼

Write an entry after any repeatable failure: bounced outreach, negative replies, copy failing quality gates, hallucinated customer data, wrong pricing, wrong brand voice, or wrong exclude lists.

What makes a good root cause in a failure log?▼

A good root cause is actionable and specific, such as a missing context file or wrong assumption, not a generic statement like the AI made a mistake. The fix must be a concrete step, not advice to be more careful.

What happens if the growth-os repo is not cloned locally?▼

If growth-os is not available, note the correction and tell the user to add it manually to growth-os/eval/corrections.md in the private repo.

Related Skills

Explore More Marketing & Sales Skills→
core-wrkcore-wrk

investor-outreach-drafter

Draft personalized investor outreach messages under 100 words.

Official
Intermediate
IgorGanapolskyIgorGanapolsky
25

recall-context

Retrieve prevention rules, past failures, and context packs via the recall MCP tool.

Community
Basic
systempromptiosystempromptio
2

brand-review

Review content against systemprompt brand identity and governance standards before publication.

Official
Intermediate
slowman2084slowman2084
6

meta-eval-judge

Score agent outputs against YAML rubrics or gold references with verified calculations.

Community
Advanced