agents-md-revise

Captures session learnings into AGENTS.md or CLAUDE.md project-rules files.

507|40|Updated Apr 26, 2026
One-click install
npx skills add https://github.com/waybarrios/opencode-power-pack --skill agents-md-revise-waybarrios
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agents-md-revise
Source: https://github.com/waybarrios/opencode-power-pack/tree/main/skills/agents-md-revise
Command: npx skills add https://github.com/waybarrios/opencode-power-pack --skill agents-md-revise-waybarrios

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Valuable context discovered during a coding session—build commands, environment quirks, project conventions—is lost when the session ends, forcing future sessions to rediscover it. This Skill reviews the current session and writes those learnings into the project-rules file so future sessions start with that knowledge. ## Core Features & Use Cases - Session Reflection: Identifies recurring, non-obvious learnings such as bash commands, testing approaches, tool versions, and environment gotchas. - Client-Aware Targeting: Resolves the correct rules file across OpenCode and Claude Code, including Claude @ imports (up to four hops), OpenCode configured sources, and nested path scopes. - Approval-Gated Writes: Shows each proposed addition as a diff and edits files only after explicit user approval, preserving existing file structure. - Use Case: After a long debugging session where you discovered the dev server binds to port 3000 and tests require a specific flag, ask the agent to save these findings to AGENTS.md so the next session avoids the same pitfalls. ## Quick Start Ask the agent to update AGENTS.md with what we learned in this session and approve the proposed diffs.

Frequently Asked Questions about agents-md-revise

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

FAQPage Schema
How do I save session learnings to AGENTS.md or CLAUDE.md?▼

Ask the agent to revise the project rules with what was learned in the session. The skill reflects on missing context, drafts concise one-line additions, shows each change as a diff, and writes only after your explicit approval.

What is the difference between agents-md-revise and agents-md-improver?▼

agents-md-improver audits existing project rules to find gaps, while agents-md-revise captures fresh learnings from the current session. They are designed to work together as a maintenance loop: improver audits, revise captures.

Does this work with both Claude Code and OpenCode rule files?▼

Yes. It resolves the effective target per client, following Claude @ imports up to four hops and OpenCode configured sources and nested scopes. It prefers a canonical AGENTS.md with a CLAUDE.md containing @AGENTS.md when both clients are required.

Can it write secrets or credentials into the rules file?▼

No. Secret values are never written to any prompt-loaded file, including local or gitignored rules. Encountered values are replaced with [REDACTED], keeping only location, type, and remediation evidence.

What happens if I reject a proposed rule addition?▼

The skill does not retry a rejected addition in the same session. It moves on to the next proposal, and only files you explicitly approve are edited.