Improving instructions

Agents.md Improver

Audit, evaluate, and improve agent-instruction files (AGENTS.md, CLAUDE.md, and variants) across a codebase to ensure the host agent has optimal project context.

This skill can write to agent-instruction files. After presenting a quality report and getting user approval, it updates files with targeted improvements.

Platform Adaptation

Capability	Claude Code	Codex
Find files	Glob / Grep	shell (find, grep)
Read a file	Read	shell (cat)
Edit a file	Edit	apply_patch / shell heredoc
User confirmation	AskUserQuestion	ask_user / built-in approval prompt
Shell commands	Bash	shell

The skill body refers to actions abstractly ("read the file", "apply the diff"). The host agent maps to its own tool inventory.

Workflow

Phase 1: Discovery

Find every agent-instruction file in scope (project + user-global), dedupe symlinked pairs via realpath:

{
  find . \( \
      -name "AGENTS.md" \
      -o -name "AGENTS.local.md" \
      -o -name "CLAUDE.md" \
      -o -name "CLAUDE.local.md" \
      -o -name ".claude.md" \
      -o -name ".claude.local.md" \
    \) -not -path '*/node_modules/*' -not -path '*/.git/*' 2>/dev/null
  ls ~/.claude/CLAUDE.md ~/.codex/AGENTS.md 2>/dev/null
} | xargs -I{} realpath {} | sort -u

File Types & Locations:

Type	Location	Purpose
Project-shared	./AGENTS.md, ./CLAUDE.md, ./.claude.md	Primary project context, in git, shared with team
Project-local	./AGENTS.local.md, ./CLAUDE.local.md, ./.claude.local.md	Personal overrides, gitignored
User-global	~/.claude/CLAUDE.md, ~/.codex/AGENTS.md	User-wide defaults across all projects
Package-specific	./packages/*/AGENTS.md, ./packages/*/CLAUDE.md	Module-level context in monorepos
Subdirectory	Any nested location	Feature/domain-specific context

Note: The host agent auto-discovers agent-instruction files in parent directories, making monorepo setups work automatically. When two paths share the same realpath (typical when CLAUDE.md is a symlink to AGENTS.md), audit them as one logical file and edit the canonical realpath target.

Phase 2: Quality Assessment

For each unique file, evaluate against quality criteria. See references/quality-criteria.md for detailed rubrics.

Scope-aware lens: the assessment lens shifts per scope. For project files: "is this codebase well-documented for an agent?". For user-global files: "are these rules organized, non-contradictory, at the right abstraction level, and not duplicated across projects?".

Quick Assessment Checklist:

Criterion	Weight	Check
Commands/workflows documented	High	Are build/test/deploy commands present?
Architecture clarity	High	Can the host agent understand the codebase structure?
Non-obvious patterns	Medium	Are gotchas and quirks documented?
Conciseness	Medium	No verbose explanations or obvious info?
Currency	High	Does it reflect current codebase state?
Actionability	High	Are instructions executable, not vague?

Quality Scores:

• A (90-100): Comprehensive, current, actionable
• B (70-89): Good coverage, minor gaps
• C (50-69): Basic info, missing key sections
• D (30-49): Sparse or outdated
• F (0-29): Missing or severely outdated

Phase 3: Quality Report Output

ALWAYS output the quality report BEFORE making any updates.

Format:

## Agents.md Quality Report

### Summary
- Files found: X (Y unique after realpath dedup)
- Average score: X/100
- Files needing update: X

### File-by-File Assessment

#### 1. ./AGENTS.md (Project-shared, also reachable as ./CLAUDE.md symlink)
**Score: XX/100 (Grade: X)**

| Criterion | Score | Notes |
|-----------|-------|-------|
| Commands/workflows | X/20 | ... |
| Architecture clarity | X/20 | ... |
| Non-obvious patterns | X/15 | ... |
| Conciseness | X/15 | ... |
| Currency | X/15 | ... |
| Actionability | X/15 | ... |

**Issues:**
- [List specific problems]

**Recommended additions:**
- [List what should be added]

#### 2. ./packages/api/AGENTS.md (Package-specific)
...

Phase 4: Targeted Updates

After outputting the quality report, ask user for confirmation before updating.

Update Guidelines (Critical):

1. Propose targeted additions only: focus on genuinely useful info:

   • Commands or workflows discovered during analysis
   • Gotchas or non-obvious patterns found in code
   • Package relationships that weren't clear
   • Testing approaches that work
   • Configuration quirks

2. Keep it minimal: avoid:

   • Restating what's obvious from the code
   • Generic best practices already covered
   • One-off fixes unlikely to recur
   • Verbose explanations when a one-liner suffices

3. Show diffs: for each change, show:

   • Which file to update (use the canonical realpath)
   • The specific addition (as a diff or quoted block)
   • Brief explanation of why this helps future sessions

Diff Format:

### Update: ./AGENTS.md

**Why:** Build command was missing, causing confusion about how to run the project.

````diff
+ ## Quick Start
+
+ ```bash
+ npm install
+ npm run dev  # Start development server on port 3000
+ ```
````

Phase 5: Apply Updates

After user approval, apply changes (see Platform Adaptation for the right edit tool). Preserve existing content structure. Edit the canonical realpath target only; symlinks update automatically.

Templates

See references/templates.md for templates by project type.

Update Guidelines Reference

See references/update-guidelines.md for full do's and don'ts.

Common Issues to Flag

1. Stale commands: build commands that no longer work
2. Missing dependencies: required tools not mentioned
3. Outdated architecture: file structure that's changed
4. Missing environment setup: required env vars or config
5. Broken test commands: test scripts that have changed
6. Undocumented gotchas: non-obvious patterns not captured
7. Scope mismatch: project-specific rules in the user-global file (or vice versa)
8. Duplicate rules: same rule stated in both project and user-global file
9. Bare prohibitions: "do not X" rules with no reason or alternative attached; propose restating as intent plus the preferred alternative
10. Constraint accretion: defensive rules whose failure mode no longer occurs; propose deletion when nobody can name what the rule prevents

User Tips to Share

When presenting recommendations, remind users:

• # key shortcut (Claude Code): during a Claude Code session, press # to have Claude auto-incorporate learnings into the project memory file
• Keep it concise: agent-instruction files should be human-readable; dense is better than verbose
• Actionable commands: all documented commands should be copy-paste ready
• Use *.local.md: for personal preferences not shared with team (add to .gitignore)
• Global defaults: put user-wide preferences in ~/.claude/CLAUDE.md or ~/.codex/AGENTS.md
• Symlink for cross-runtime parity: if you use both Claude Code and Codex, symlink CLAUDE.md → AGENTS.md so a single file serves both

What Makes a Great Agent-Instruction File

Key principles:

• Concise and human-readable
• Actionable commands that can be copy-pasted
• Project-specific patterns, not generic advice
• Non-obvious gotchas and warnings

Recommended sections (use only what's relevant):

• Commands (build, test, dev, lint)
• Architecture (directory structure)
• Key Files (entry points, config)
• Code Style (project conventions)
• Environment (required vars, setup)
• Testing (commands, patterns)
• Gotchas (quirks, common mistakes)
• Workflow (when to do what)