meta-dev
Skill17 starsChanged 4 months agoMITView on GitHub
AGENT-11 framework development. Use when modifying agents, commands, missions, skills, or library files inside the agent-11 repository itself. Distinguishes the LIBRARY surface (project/agents/specialists/, project/commands/, library/) — which is deployed to users — from the WORKING SQUAD (.claude/agents/, .claude/commands/) — which is internal-only dev tooling. Triggered by the AGENT_11_PROJECT env var set in .claude/settings.local.json.
In this repository10 other files it gives its agents.
- .claude/claude.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- library/CLAUDE.mdCLAUDE.md
- saas-analyticsSkill
- saas-authSkill
- saas-billingSkill
- saas-emailSkill
- saas-multitenancySkill
- saas-onboardingSkill
- saas-paymentsSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 47
When in doubt: project/agents/specialists/ and project/commands/.
- Distinctive: line 61
Before editing an agent or command file, ask:
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
AGENT-11 Meta-Development
What this skill is for
You are working inside the AGENT-11 framework repository, not a user product. Edits here ship to other people via install.sh. The same filename can mean two different things depending on directory.
This skill orients you correctly without relying on memory or repeated reminders.
Two parallel agent sets in this repo
| Path | Role | Modify when |
project/agents/specialists/ | Library agents — deployed to users via install.sh | You are improving agent capabilities for end users (≥99% of work) |
.claude/agents/ | Working squad — internal dev tooling | You are improving how we build AGENT-11 itself (rare) |
Same distinction for commands:
| Path | Role | Modify when |
project/commands/ | Library commands — deployed | Improving end-user /coord, /foundations, etc. |
.claude/commands/ | Working squad commands — internal | Improving internal dev workflows |
Default work target
When in doubt: project/agents/specialists/ and project/commands/. Most work improves what users get on install.sh, not our internal tooling.
CLAUDE.md hierarchy in this repo
| File | Purpose | Deployed? |
/CLAUDE.md | Jamie's personal preferences | NO |
library/CLAUDE.md | AGENT-11 library instructions → user's .claude/CLAUDE.md post-install | YES |
.claude/CLAUDE.md | Working-squad project guardrails | NO |
Edits to library content target library/CLAUDE.md. Personal preferences belong in root. Internal dev guardrails belong in .claude/CLAUDE.md.
Verification before significant edits
Before editing an agent or command file, ask:
- Am I improving the AGENT-11 library for end users? →
project/agents/specialists/orproject/commands/ - Am I improving the AGENT-11 development process? →
.claude/agents/or.claude/commands/ - Quick check:
grep -n "$(basename your-file)" project/deployment/scripts/install.sh— if found, it's deployed (library); if not, it's internal.
Sprint 4 v6.0 evolution scope rule
During v6.0 sprints (4a–4h), edits target the library surface only:
project/agents/specialists/project/commands/project/missions/project/field-manual/project/skills/project/deployment/library/CLAUDE.mdlibrary/settings.json.templatetemplates/
The working squad (.claude/agents/, .claude/commands/) is NOT modified during v6.0 — it is the squad we use to do the v6.0 work. The one explicit exception per the v6.0 plan: .claude/skills/meta-dev/ (this file).
Anti-patterns to catch
- Editing
.claude/agents/coordinator.mdthinking you're improving the deployed coordinator → wrong file. Library coordinator isproject/agents/specialists/coordinator.md. - Editing
.claude/CLAUDE.mdthinking you're updating user-facing instructions → wrong file. User-facing islibrary/CLAUDE.md. - Adding install steps without updating
install.sh's SHA256 (install.sh.sha256) → CI guard will fail. - Modifying agents without updating their schema validation if they have one → pre-commit hook will block.
Reference
.claude/CLAUDE.md carries the full distinction in detail. This skill is a routing aid — use it to default-orient, then read the working-squad file when you need the longer explanation.
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.

