writing-project-constitution
Skill2 starsChanged 8 months agoMITView on GitHub
Creates project constitution files (CLAUDE.md/AGENTS.md) that serve as always-loaded context for coding agents. Use when setting up a new project for spec-driven development, configuring agent instructions, writing CLAUDE.md or AGENTS.md, or establishing project-wide coding standards and constraints.
- Reads credentials
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- building-lit-componentsSkill
- building-static-site-rss-feedsSkill
- building-templ-componentsSkill
- checking-wsh-vrtSkill
- committing-codeSkill
- creating-agent-skillsSkill
- creating-test-casesSkill
- deploying-with-isucon-ansibleSkill
- designing-isucon-architectureSkill
- frontend-designSkill
- operating-isucon-servers-via-sshSkill
- planning-wsh-improvementsSkill
- porting-isucon-rust-to-goSkill
- releasing-zig-with-goreleaserSkill
- running-isucon-as-an-agentSkill
- sharing-sockets-with-so-reuseport-in-zigSkill
- tuning-isucon-linux-and-nginxSkill
- using-go-tool-directiveSkill
- using-grafana-mcpSkill
- using-unocss-with-templSkill
- winning-isuconSkill
- winning-web-speed-hackathonSkill
- writing-feature-specSkill
- writing-implementation-tasksSkill
- writing-isucon-arch-subagent-tasksSkill
- writing-technical-designSkill
- writing-zig-cli-toolsSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Create constitution files (CLAUDE.md / AGENTS.md) that are always loaded into the agent's context window on every request.
- Distinctive: line 115
Every line in the constitution costs tokens on every single request.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Writing Project Constitution
Create constitution files (CLAUDE.md / AGENTS.md) that are always loaded into the agent's context window on every request. This is the most expensive context layer -- every token here is consumed on every interaction.
Use this skill when initializing a project for spec-driven development, writing CLAUDE.md or AGENTS.md, or defining always-applicable project rules.
Supporting files: CROSS-AGENT.md for agent-specific file placement, EXAMPLES.md for complete examples.
Context Layer: L1 (Always Loaded)
Context Loading Timeline
================================================================
Session Start Task Start Implementation Deep Dive
|
+-- CLAUDE.md / AGENTS.md <-- THIS SKILL
| (loaded EVERY request, costs tokens EVERY time)
|
+-- Skill descriptions (metadata only)
================================================================The constitution is the only document that persists across context compression. When the agent's conversation gets too long, older messages are compressed, but the constitution is re-injected. This makes it the right place for information that must NEVER be forgotten.
What Belongs Here (and What Does NOT)
MUST include (needed every request):
- Build / test / lint commands
- Tech stack summary (one line per technology)
- Core coding conventions (naming, error handling style)
- Hard constraints ("never commit secrets", "always run tests")
- Active feature spec references (one line each)
- Directory structure overview (abbreviated)
MUST NOT include (move elsewhere):
| Information | Move to | Reason |
| Detailed requirements | Feature spec (L3/L4) | Only needed during implementation |
| Architecture decisions | Technical design (L3/L4) | Only needed during design |
| API endpoint details | Path-conditional rules (L2) | Only needed for specific files |
| Task lists | Implementation tasks (L3) | Only needed when selecting work |
| Code examples > 5 lines | Agent Skill references (L4) | Too expensive for L1 |
| Library documentation | MCP or references (L4) | Agent can look up on demand |
Template
# Project Constitution
## Commands
- Build: `[command]`
- Test: `[command]`
- Test single: `[command with placeholder]`
- Lint: `[command]`
- Format: `[command]`
## Tech Stack
- Language: [e.g., Go 1.23]
- Framework: [e.g., Echo v4]
- Database: [e.g., PostgreSQL 16 via sqlc]
- Frontend: [e.g., templ + htmx]
## Project Structure
src/ # Application source internal/ # Private packages cmd/ # Entry points specs/ # Feature specs and designs (see below)
## Coding Standards
- [Convention 1: e.g., Use snake_case for DB columns, camelCase for Go]
- [Convention 2: e.g., Errors must wrap with fmt.Errorf("context: %w", err)]
- [Convention 3: e.g., All public functions must have godoc comments]
## Boundaries
- ALWAYS: Run tests before considering implementation complete
- ALWAYS: Follow existing patterns in the codebase
- ASK FIRST: Database schema changes, new dependencies
- NEVER: Commit .env files, modify CI config, use force push
## Active Specs
Current feature specs (load these when working on related code):
- `specs/prd-notifications.md` - Push notification system
- `specs/design-notifications.md` - Notification architecture
- `specs/tasks-notifications.md` - Implementation tasks
## Conditional Rules
Code-area-specific rules are in:
- `.claude/rules/` (Claude Code)
- `.github/instructions/` (Copilot CLI)
These load automatically when touching matching files.Size Budget
Target: under 200 lines / ~2000 tokens.
Every line in the constitution costs tokens on every single request. Measure ruthlessly:
Good constitution (150 lines):
- 10 lines: commands
- 5 lines: tech stack
- 10 lines: project structure
- 15 lines: coding standards
- 10 lines: boundaries
- 10 lines: active specs
- Remaining: breathing room
Bad constitution (500+ lines):
- 50 lines: detailed API documentation -> move to L4
- 100 lines: code examples -> move to L4
- 80 lines: architecture decisions -> move to L3/L4
- 200 lines: task lists -> move to L3Cross-Agent File Placement
The same content maps to different files per agent:
| Agent | Primary File | Location |
| Claude Code | CLAUDE.md | Project root |
| Codex CLI | AGENTS.md | Project root |
| Copilot CLI | AGENTS.md + .github/copilot-instructions.md | Project root |
| Multi-agent | Both CLAUDE.md and AGENTS.md | Project root |
For multi-agent projects, create both files. They can share content or reference each other. See CROSS-AGENT.md for detailed patterns.
Path-Conditional Rules (L2 Layer)
Move area-specific rules out of the constitution into conditional files:
Claude Code: .claude/rules/
<!-- .claude/rules/api-handlers.md -->
---
paths:
- "internal/handler/**/*.go"
- "internal/api/**/*.go"
---
API handler conventions:
- Use echo.Context for HTTP handling
- Return HTTP errors via echo.NewHTTPError()
- Validate request body with validator tags
- Log with structured logger (slog)Copilot CLI: .github/instructions/
<!-- .github/instructions/api-handlers.instructions.md -->
---
applyTo: "internal/handler/**/*.go,internal/api/**/*.go"
---
API handler conventions:
- Use echo.Context for HTTP handling
...Codex CLI: Subdirectory AGENTS.md
<!-- internal/handler/AGENTS.md -->
API handler conventions:
- Use echo.Context for HTTP handling
...Integration with Other Spec Documents
The constitution is the hub that links to other documents. Keep references minimal:
## Active Specs
- `specs/prd-{feature}.md` - What to build (load via feature-spec skill)
- `specs/design-{feature}.md` - How to build (load via technical-design skill)
- `specs/tasks-{feature}.md` - Implementation order (load via implementation-tasks skill)Agents use these references to know which files to load when starting work on a feature.
Workflow
1. Create CLAUDE.md (and/or AGENTS.md) at project root
2. Add build/test commands first (most universally needed)
3. Add tech stack summary (one line per technology)
4. Add coding standards (only project-specific ones)
5. Add boundaries (always/ask-first/never)
6. Extract area-specific rules to L2 conditional files
7. Add active spec references as features are planned
8. Review total line count -- cut anything over 200 lines
9. Validate: read the file and ask "is EVERY line needed on EVERY request?"Quality Checklist
Constitution Quality Check:
- [ ] Under 200 lines total
- [ ] Build/test/lint commands are present and correct
- [ ] No detailed requirements (those belong in feature specs)
- [ ] No architecture decisions (those belong in technical designs)
- [ ] No task lists (those belong in implementation tasks)
- [ ] No code examples longer than 5 lines
- [ ] Area-specific rules extracted to L2 conditional files
- [ ] Active spec references are current (no stale links)
- [ ] Every line passes the test: "needed on EVERY request?"Detailed Guides
Cross-agent file placement: See CROSS-AGENT.md Complete constitution examples: See EXAMPLES.md
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

