# Agent Context Wiring > Wire standards into CLAUDE.md, AGENTS.md, and .cursor/rules/ using references Use when this capability is needed. - Skill: `tomevault-io/agent-context-wiring` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/agent-context-wiring` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/agent-context-wiring/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/agent-context-wiring --- # Agent Context Wiring Skill Ensure that standards are discoverable by agent files without duplicating content. ## Context Agent files (CLAUDE.md, AGENTS.md, .cursor/rules) guide Claude and other developers through your standards. These files must reference standards — never duplicate them — so that updates to a standard are immediately visible to agents. ## Domain Context The key principle of claude-standards is **reference, never duplicate**. - Source of truth: `docs/standards/` contains standard documents - Agent files: CLAUDE.md, AGENTS.md, .cursor/rules/ contain references to standards - When a standard is updated, agents automatically see the new version - Agent files are always kept in sync with reality Reference: `levels/L1-context/templates/` for example context files ## Instructions 1. **Identify which agent tools are in use** — Determine which files guide your team - Does your project use CLAUDE.md? (Almost always yes) - Does your project use AGENTS.md? (Yes, if you have team processes) - Does your project use .cursor/rules/? (Optional, for IDE context) - Deliverable: Inventory of context files in your project 2. **Create or update CLAUDE.md** — Add standards references - Should reference the standards directory: `docs/standards/` - Should list applicable standards by category - Example entry: `See docs/standards/engineering/typescript.md for type safety rules` - Rule: Always use relative paths (`./docs/standards/...`) - Deliverable: Updated CLAUDE.md with standards references 3. **Create or update AGENTS.md** — Add team process guidance - Should explain which standards agents should focus on - Should reference ceremonies and enforcement processes - Example: `Before writing code, agents should read docs/standards/engineering/` - Rule: Always use relative paths - Deliverable: Updated AGENTS.md with standards references 4. **Create .cursor/rules/ files** (if applicable) — Add IDE-level rules - Can create separate rule files for different categories - Example: `.cursor/rules/engineering.md` references `docs/standards/engineering/` - Use consistent format across rule files - Deliverable: New or updated .cursor/rules/ files 5. **Verify standards are referenced, not duplicated** — Check for content duplication - Search for exact phrases from standards in agent files - If found, replace with reference: `See docs/standards/[category]/[standard].md` - Guard: Agent files should never contain more than 1-2 line summaries - Deliverable: Agent files contain only references, no duplicated standard content 6. **Run check-standards-refs.sh** — Verify all references are valid ```bash ./levels/L2-ceremonies/scripts/check-standards-refs.sh ``` - This script finds broken references (standards that don't exist) - Fix any broken paths before merging - Deliverable: All verification scripts pass ## Anti-Patterns 1. **Duplicating standard content into context files** — Major maintenance burden - Wrong: Copy the entire error-handling standard into AGENTS.md - Right: "See docs/standards/engineering/error-handling.md for patterns" - Impact: When standard updates, agent file becomes stale - Guard: If you find yourself copy-pasting from a standard, stop and add a reference instead 2. **Missing wiring** — Standard exists but no agent sees it - Wrong: Write a new standard but forget to mention it in CLAUDE.md - Right: Create standard, add to README index, reference in CLAUDE.md - Impact: Standard is written but never discovered - Guard: Before marking a standard "done", verify it's referenced in at least one agent file 3. **Hardcoded file paths instead of relative paths** — Breaks in different contexts - Wrong: `/Users/someone/projects/repo/docs/standards/engineering/api.md` - Right: `./docs/standards/engineering/api.md` - Impact: Paths break when repo is cloned or moved - Guard: All paths should be relative to project root 4. **Inconsistent reference format** — Agent files don't follow a pattern - Wrong: Some files use `See docs/...`, some use `Read about...`, some use `Check...` - Right: Use consistent format: `See [Standard Title](./docs/standards/category/file.md)` - Guard: Use markdown link format for consistency: `[Title](path)` ## Further Reading - `levels/L1-context/templates/` — Example CLAUDE.md and AGENTS.md - `levels/L2-ceremonies/scripts/check-standards-refs.sh` — Reference verification script - `AGENTS.md` (in this repo) — Example of well-wired agent context --- > Source: [sethdford/claude-standards](https://github.com/sethdford/claude-standards) — distributed by [TomeVault](https://tomevault.io).