Agent skill
maintaining-project-context
Use when completing development phases or branches to identify and update CLAUDE.md or AGENTS.md files that may have become stale - analyzes what changed, determines affected contracts and documentation, and coordinates updates
Install this agent skill to your Project
npx add-skill https://github.com/ed3dai/ed3d-plugins/tree/main/plugins/ed3d-extending-claude/skills/maintaining-project-context
SKILL.md
Maintaining Project Context
REQUIRED SUB-SKILL: Use ed3d-extending-claude:writing-claude-md-files for all context file creation and updates.
Core Principle
Context files (CLAUDE.md or AGENTS.md) document contracts and architectural intent. When code changes contracts, the documentation must update. Stale documentation is worse than no documentation.
Trigger: End of development phase, branch completion, or any work that changed contracts, APIs, or domain structure.
Format Detection (MANDATORY FIRST STEP)
Before any updates, detect what format this repository uses:
# Check for AGENTS.md at root
ls -la AGENTS.md 2>/dev/null
# Check for CLAUDE.md at root
ls -la CLAUDE.md 2>/dev/null
| Root AGENTS.md? | Format | Action |
|---|---|---|
| Yes | AGENTS.md-canonical | Update AGENTS.md files, create companion CLAUDE.md |
| No | CLAUDE.md-canonical | Update CLAUDE.md files directly |
Key principle: We use OUR format structure (Purpose, Contracts, Dependencies, Invariants, etc.) regardless of filename. AGENTS.md is just for cross-platform AI agent compatibility.
AGENTS.md-Canonical Repos
When the repo uses AGENTS.md:
- Read AGENTS.md first before making any updates
- Write content to AGENTS.md using our standard structure
- Create companion CLAUDE.md next to each AGENTS.md with exactly this content:
Read @./AGENTS.md and treat its contents as if they were in CLAUDE.md
When to Update Context Files
| Change Type | Update Required? | What to Update |
|---|---|---|
| New domain/module | Yes | Create domain context file |
| API/interface change | Yes | Contracts section |
| Architectural decision | Yes | Key Decisions section |
| Invariant change | Yes | Invariants section |
| Dependency change | Yes | Dependencies section |
| Bug fix (no contract change) | No | - |
| Refactor (same behavior) | No | - |
| Test additions | No | - |
The Process
Step 1: Identify What Changed
Diff against the base (branch start or phase start):
# Get changed files
git diff --name-only <base-sha> HEAD
# Get detailed changes
git diff <base-sha> HEAD --stat
Categorize changes:
- Structural: New directories, moved files
- Contract: Changed exports, interfaces, public APIs
- Behavioral: Changed invariants, guarantees
- Internal: Implementation details only
Step 2: Map Changes to Context Files
For each significant change, determine which context file should document it:
| Change Location | Context File Location |
|---|---|
| Project-wide pattern | Root context file |
| New domain | <domain>/ context file (create) |
| Existing domain contract | <domain>/ context file (update) |
| Cross-domain dependency | Both affected domains |
Hierarchy rule: Information belongs at the lowest level where it applies. Domain-specific contracts go in domain files, not root.
For AGENTS.md-canonical repos: When creating new domain context files, create both AGENTS.md (with content) and CLAUDE.md (companion pointer).
Step 3: Verify Contracts Still Hold
For each affected context file, verify:
- Contracts section: Do exposes/guarantees/expects match current code?
- Dependencies section: Are uses/used-by/boundary accurate?
- Invariants section: Are all invariants still enforced?
- Key Decisions section: Any new decisions to document?
# Find domain's public exports
grep -r "export" <domain>/index.ts
# Find domain's imports (dependencies)
grep -r "from '\.\." <domain>/
Step 4: Update or Create Context Files
For updates:
- Read existing file first (especially for AGENTS.md)
- Update freshness date via
date +%Y-%m-%d - Update affected sections
- Remove stale content
- Verify under token budget (<100 lines for domain files)
For new domains (CLAUDE.md-canonical repos):
- Create
<domain>/CLAUDE.mdusing template from writing-claude-md-files - Document purpose, contracts, dependencies, invariants
- Set freshness date
For new domains (AGENTS.md-canonical repos):
- Create
<domain>/AGENTS.mdusing template from writing-claude-md-files - Document purpose, contracts, dependencies, invariants
- Set freshness date
- Create companion
<domain>/CLAUDE.md:markdownRead @./AGENTS.md and treat its contents as if they were in CLAUDE.md
Step 5: Commit Documentation Updates
git add <affected CLAUDE.md files>
git commit -m "docs: update project context for <branch-name>"
Decision Tree
Has code changed?
├─ No → Skip (nothing to update)
└─ Yes → Detect format first (AGENTS.md at root?)
│
└─ What changed?
├─ Only tests/internal details → Skip
└─ Contracts/APIs/structure → Continue
│
├─ New domain created?
│ ├─ AGENTS.md repo → Create AGENTS.md + companion CLAUDE.md
│ └─ CLAUDE.md repo → Create CLAUDE.md
│
├─ Existing domain changed?
│ └─ Update domain context file (read first!)
│
└─ Project-wide pattern changed?
└─ Update root context file
Quick Reference
Always update when:
- New public exports added
- Interface signatures changed
- Invariants added/removed
- Dependencies changed
- Architectural decisions made
Never update for:
- Internal refactoring
- Bug fixes that don't change contracts
- Test file changes
- Comment/documentation-only changes
Common Mistakes
| Mistake | Fix |
|---|---|
| Updating for every change | Only update for contract changes |
| Forgetting freshness date | Always use date +%Y-%m-%d |
| Documenting implementation | Document contracts and intent |
| Putting domain info in root | Use domain context files for domain contracts |
| Skipping verification | Read the code, confirm contracts hold |
| Skipping format detection | Always check for AGENTS.md first |
| Writing AGENTS.md without reading | Always read existing content before updating |
| Forgetting companion CLAUDE.md | AGENTS.md repos need both files |
Integration Points
Called by:
- project-claude-librarian agent - Uses this skill to coordinate updates
- executing-an-implementation-plan (Step 5b) - After all tasks complete
- finishing-a-development-branch (Step 4b) - Before merge/PR
Uses:
- writing-claude-md-files - For actual context file creation/updates (works for both CLAUDE.md and AGENTS.md)
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
doing-a-simple-two-stage-fanout
Use when analyzing a large corpus of text, code, or data that exceeds a single agent's effective context - orchestrates parallel Worker subagents, Critic review subagents, and a final Summarizer subagent with task tracking and failure recovery
using-generic-agents
Use to decide what kind of generic agent you should use
investigating-a-codebase
Use when planning or designing features and need to understand current codebase state, find existing patterns, or verify assumptions about what exists; when design makes assumptions about file locations, structure, or existing code that need verification - prevents hallucination by grounding plans in reality
researching-on-the-internet
Use when planning features and need current API docs, library patterns, or external knowledge; when testing hypotheses about technology choices or claims; when verifying assumptions before design decisions - gathers well-sourced, current information from the internet to inform technical decisions
creating-an-agent
Use when creating specialized subagents for Claude Code plugins or the Task tool - covers description writing for auto-delegation, tool selection, prompt structure, and testing agents
prompt-security-hardening
Use when writing skills, CLAUDE.md files, agent prompts, or any directives that involve shell commands, environment variables, API credentials, file creation, or git operations - prevents secrets leakage into LLM context, unsafe shell patterns, and credential exposure
Didn't find tool you were looking for?