a-team / rules
AGENTS.md24 starsChanged 24 days agoMITView on GitHub
Use parallel dispatch for independent operations: A Team uses three tiers regardless of platform. The model: values in agent frontmatter are Claude Code defaults — on other platforms, select the equivalent tier in your platform's model settings. Use the lowest tier that can handle the task to control cost. For complex decisions, use split role sub-agents in parallel: - Factual reviewer - Senior engineer perspective - Security expert perspective - Consistency reviewer Avoid the last 20% of context window for:…
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 repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- api-contract-firstSkill
- architecture-auditSkill
- brainstormingSkill
- data-migrationSkill
- dispatching-parallel-agentsSkill
- executing-plansSkill
- finishing-a-development-branchSkill
- five-whysSkill
- incident-responseSkill
- performance-auditSkill
- skill-duplication-auditSkill
- smart-initSkill
- subagent-driven-developmentSkill
- systematic-debuggingSkill
- test-driven-developmentSkill
- using-a-teamSkill
- using-git-worktreesSkill
- verification-before-completionSkill
- writing-plansSkill
- writing-skillsSkill
The file
We split this agents.md into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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 17
Use parallel dispatch for independent operations:
- Distinctive: line 40
Use the lowest tier that can handle the task to control cost.
- Distinctive: line 60
Avoid the last 20% of context window for:
- Distinctive: line 65
Prefer fresh sub-agents for isolated tasks to preserve coordinator context.
≈ “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 Rules
Immediate Agent Triggers (No User Prompt Needed)
| Situation | Agent |
| Complex feature request | planner |
| Code just written/modified | code-reviewer |
| Bug reported or test failing | debugger |
| Build broken | build-error-resolver |
| Architectural decision needed | architect |
| Security-sensitive code | security-reviewer |
| New feature (full flow) | brainstorming → planner → subagent-driven-development |
Parallel Execution (ALWAYS)
Use parallel dispatch for independent operations:
# CORRECT: Parallel
Launch in parallel:
1. security-reviewer on auth module
2. code-reviewer on cache system
3. go-reviewer on utilities
# WRONG: Sequential when unnecessary
First security, then code, then go-reviewerModel Tier Selection
A Team uses three tiers regardless of platform. The model: values in agent frontmatter are Claude Code defaults — on other platforms, select the equivalent tier in your platform's model settings.
| Tier | Capability | Claude Code | OpenAI / Codex | Google / Gemini | Use for |
| Tier 1 | Complex reasoning | claude-opus-4-8 | o3 / gpt-5.5 | gemini-2.5-pro | Orchestrator, Architect |
| Tier 2 | Balanced | claude-sonnet-4-6 | gpt-4.1 | gemini-2.5-flash | Most agents — review, debug, implement |
| Tier 3 | Lightweight | claude-haiku-4-5 | gpt-4.1-mini | gemini-2.0-flash-lite | Docs, audit, profiling |
Use the lowest tier that can handle the task to control cost.
Multi-Perspective Analysis
For complex decisions, use split role sub-agents in parallel:
- Factual reviewer
- Senior engineer perspective
- Security expert perspective
- Consistency reviewer
Agent Scope Rules
- Agents must not exceed their stated tool permissions
build-error-resolvermakes minimal diffs only — no refactoringdebuggermust complete Phase 1 before proposing any fixorchestratornever writes product coderefactor-cleanernever runs during active feature development
Context Window Management
Avoid the last 20% of context window for:
- Large-scale refactoring
- Multi-file feature implementation
- Complex debugging sessions
Prefer fresh sub-agents for isolated tasks to preserve coordinator context.
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.