6SNT-Studio
AGENTS.md0 starsChanged 5 months agoMITView on GitHub
Ten specialized agents plus a non-producing Lead. **Hard rule: an agent only writes its own deliverable.** Any agent may read what it needs; to write, only its zone. This reinforces "don't modify approved artifacts from earlier stages" and "boundary clashes are escalated, not resolved between agents."
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,843 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository20 other files it gives its agents.
- adopt-groundSkill
- agent-authoringSkill
- architecture-decisionSkill
- asset-sourcingSkill
- brand-evaluation-rubricSkill
- criticSkill
- data-integritySkill
- diagrams-gateSkill
- drift-protocolSkill
- env-reproducibleSkill
- orchestrator-bridgeSkill
- project-cleanupSkill
- qa-eyesSkill
- qa-visual-criteriaSkill
- research-methodologySkill
- security-auditSkill
- setupSkill
- smoke-testSkill
- technical-documentationSkill
- visualizerSkill
The file
We split this AGENTS.md 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 23
Frontend ∥ Backend: run in parallel, isolated by the ownership map (06 owns the presentation layer.
- Distinctive: line 42
Every agent closes by delivering its artifact and reporting what it did and what's pending.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.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
Agents
Ten specialized agents plus a non-producing Lead. Hard rule: an agent only writes its own deliverable. Any agent may read what it needs; to write, only its zone. This reinforces "don't modify approved artifacts from earlier stages" and "boundary clashes are escalated, not resolved between agents."
Ownership map
| # | Agent | Model | Writes — and ONLY this |
| 01 | Strategist | opus | BRIEF.md |
| 02 | Researcher | sonnet | RESEARCH.md |
| 03 | UX/UI | sonnet | BRANDBOOK.md + TOKENS.css + previews |
| 04 | Architect | opus | ARCHITECTURE.md + src/shared/types.ts |
| 05 | Data Modeler | opus | DATAMODEL.md |
| 06 | Frontend | sonnet | src/renderer/{ui,render,state} |
| 07 | Backend | sonnet | src/main · src/core · src/preload (+ packaging the runnable artifact) |
| 08 | QA | sonnet | tests/ + QA_REPORT.md (+ authors & runs the smoke test) |
| 09 | Security | opus | SECURITY_REPORT.md — read-only over everything else |
| 10 | Technical Writer | haiku | README.md + user docs |
| — | Lead | — | nothing — only coordinates |
Boundary notes
- 09 Security is read-only: it never writes, runs, or modifies code or the app; its only writable output is its own report.
- Frontend ∥ Backend run in parallel, isolated by the ownership map (06 owns the presentation layer; 07 owns main/core/preload). The integration contract is defined first; integration doubts are escalated to the lead, not edited across the boundary.
- 07 owns packaging: shipping the executable artifact (e.g. portable + installer) is part of its "done", not a post-release afterthought.
- Adversarial pre-gate (
critic): 02, 04, 08 and 09 run thecriticskill on their own deliverable before its gate (external review beats self-review). 04 uses it for cross-artifact consistency at G04 (does ARCHITECTURE honor BRIEF / RESEARCH / BRANDBOOK?).
Escalation topology
AGENT (01–10)
│ conflict between agents, ambiguity, or a doubt with
│ design/architecture/data implications
▼
LEAD ──pauses the pipeline──► consults the human Product Owner
│ PO decides / approves the gate
▼
lead resumes & re-activates the agent- An agent never resolves a conflict with another agent — it escalates to the lead.
- The lead never substitutes the PO: on gate decisions and design/architecture/data trade-offs it pauses and asks.
- Every agent closes by delivering its artifact and reporting what it did and what's pending.
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.