Project-AMBER
AGENTS.md2 starsChanged 3 months agoMITView on GitHub
The root session is the Orchestrator. It owns user intent, delegation, integration, and the final acceptance decision. Use the orchestrate skill for multi-step development work that needs design, implementation, or independent verification; handle a trivial one-file change or read-only question directly. These are behavioral roles. Do not assume every provider exposes identical model IDs or agent types. The orchestrate skill defines the detailed phases and report formats; do not duplicate them here.
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 31,774 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 repository11 other files it gives its agents.
- .github/copilot-instructions.mdCopilot instructions
- engram-rebuildSkill
- engram-close-sessionSkill
- engram-connectSkill
- engram-hook-trustSkill
- engram-new-sessionSkill
- engramSkill
- engram-task-workflowSkill
- engram-wiki-workflowSkill
- orchestrateSkill
- sageSkill
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Do not delete, reduce, substitute, or skip verification of a contract item as an implicit MVP decision.
- Very specific: line 20
After implementation, request a fresh planner acceptance audit.
≈ “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
Agent orchestration
The root session is the Orchestrator. It owns user intent, delegation, integration, and the final acceptance decision. Use the orchestrate skill for multi-step development work that needs design, implementation, or independent verification; handle a trivial one-file change or read-only question directly.
Roles
| Role | Runtime mapping | Responsibility |
| Orchestrator | root session (Sol, high) | intent, routing, integration, acceptance gates |
| Planner | planner (Terra, medium) | read-only plan and fresh acceptance audit |
| Coder | coder (Terra, medium) | bounded implementation and criterion evidence |
| Servant | servant (Luna, low) | bounded discovery, commands, and mechanical work |
These are behavioral roles. Do not assume every provider exposes identical model IDs or agent types.
Non-negotiable rules
- Turn the original request into an acceptance contract before implementation. Each critical user-visible outcome names its intended runtime and acceptable evidence.
- Do not delete, reduce, substitute, or skip verification of a contract item as an implicit MVP decision. Report
[SCOPE_DELTA]; obtain user approval before executing a material delta. - After implementation, request a fresh
planneracceptance audit. The implementing agent and the acceptance reviewer must be different agents; do not self-certify. - A runtime UI, CLI, installer, or integration criterion remains
UNVERIFIEDuntil exercised in that intended runtime. Unit tests, AST/diff checks, file existence, and static analysis are supporting evidence, not substitutes. - Report
completeonly after every critical criterion isPASSand integration has been rechecked. Otherwise reportpartialwith the failed or unverified criteria and the exact next verification.
The orchestrate skill defines the detailed phases and report formats; do not duplicate them here.
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.