agent-os / templates
CLAUDE.md0 starsChanged 9 months agoMITView on GitHub
When I mention these topics, automatically use the corresponding capability without asking. Recognize these implicit workload indicators: Vault location: <!-- VAULT_PATH from agent-os.config.json --> This vault is Claude's persistent memory across sessions. Use direct file access (Read, Write, Edit, Glob, Grep). At Session Start: 1. Check Sessions/ for recent handoff notes relevant to current work 2. Read Projects/<name>/README.md for context 3. Consult BRAIN.md if unsure how to use the vault During Work: - Document important decisions, findings, and context as…
In this repository9 other files it gives its agents.
- audit-instructionsSkill
- index-codebaseSkill
- meeting-notesSkill
- project-statusSkill
- red-teamSkill
- retroSkill
- session-cleanupSkill
- session-handoffSkill
- session-startSkill
The file
We split this CLAUDE.md into the rules it gives an agent
10 rules. About 0 are common, 5 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 5
When I mention these topics, automatically use the corresponding capability without asking.
- Very specific: line 42
Error self-correction: After 3+ same-type tool errors in a session, proactively add the pattern to error-patterns.md (don't wait for /retro)
- Distinctive: line 43
Context-drift detection: If MEMORY.md or vault info contradicts what you see in the current codebase, flag it to the user immediately
- Very specific: line 44
Post-work reflection nudge: After substantial work (multiple file changes, debugging sessions, or new features), suggest /retro once per …
- Very specific: line 63
Check Sessions/ for recent handoff notes relevant to current work
- Very specific: line 64
Read Projects/<name>/README.md for context
- Distinctive: line 68
Document important decisions, findings, and context as you go
- Distinctive: line 71
Update project README with any significant changes
- Very specific: line 72
Create Sessions/YYYY-MM-DD-topic.md for work-in-progress handoffs
- Distinctive: line 73
Update MEMORY.md with any new environment facts or tool discoveries
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md 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
Agent OS — Claude Code Configuration
AUTO-DISPATCH RULES
When I mention these topics, automatically use the corresponding capability without asking.
Session & Project Management
| I mention... | Auto-invoke |
| Session start, catch me up, where were we | /session-start |
| Session handoff, pause work, save context | /session-handoff |
| Cleanup, wrap up, end session | /session-cleanup |
| Retro, reflect, what did we learn | /retro |
| Project status, where are we, what's blocked | /project-status |
| Audit instructions, check rules | /audit-instructions |
| Index codebase, map codebase | /index-codebase |
| Red team this, challenge this, poke holes | /red-team |
| Meeting notes, process meeting, action items | /meeting-notes |
Knowledge & Memory
| I mention... | Auto-use |
| Remember this, save for later | Write to vault or memory files |
| What do you know about, recall | Read from vault + memory files |
<!-- Add your own auto-dispatch rules here. Examples:
| I mention... | Auto-use |
|--------------|----------|
| Calendar, schedule, meeting | `gcalcli` via Bash |
| GitHub repo, PR, issues | `gh` CLI via Bash |
| Library docs, latest docs | Context7 MCP |
-->BEHAVIOR PREFERENCES
- Auto-invoke skills — Don't ask "would you like me to use X?" — just use the right tool
- Use CLIs directly — When a CLI tool can do the job, use it immediately
- Combine capabilities — Use multiple tools together for complex tasks
SELF-IMPROVEMENT RULES
- Error self-correction: After 3+ same-type tool errors in a session, proactively add the pattern to
error-patterns.md(don't wait for/retro) - Context-drift detection: If MEMORY.md or vault info contradicts what you see in the current codebase, flag it to the user immediately
- Post-work reflection nudge: After substantial work (multiple file changes, debugging sessions, or new features), suggest
/retroonce per session
TASK INTENSITY SIGNALS
Recognize these implicit workload indicators:
| Signal Words | Intensity | Response Style |
| "quick check", "glance", "wrap up" | Light | Fast, minimal output |
| "review", "status", "sync" | Medium | Thorough but focused |
| "deep dive", "investigate", "audit" | Heavy | Comprehensive analysis |
VAULT (Agent Memory)
Vault location: <!-- VAULT_PATH from agent-os.config.json -->
This vault is Claude's persistent memory across sessions. Use direct file access (Read, Write, Edit, Glob, Grep).
At Session Start:
- Check
Sessions/for recent handoff notes relevant to current work - Read
Projects/<name>/README.mdfor context - Consult
BRAIN.mdif unsure how to use the vault
During Work:
- Document important decisions, findings, and context as you go
At Session End:
- Update project README with any significant changes
- Create
Sessions/YYYY-MM-DD-topic.mdfor work-in-progress handoffs - Update MEMORY.md with any new environment facts or tool discoveries
PROJECT-CODEBASE MAPPING
<!-- Add your projects here. This table maps vault folders to codebase paths. --><!-- The session-context-preloader hook uses agent-os.config.json for the same mapping. -->| Vault Folder | Codebase |
Projects/example-app/ | /path/to/example-app/ |
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.