pavel-molyanovpavel-molyanov/molyanov-ai-dev/AGENTS.md

molyanov-ai-dev

AGENTS.md299 starsChanged 11 days agoMITView on GitHub

- The user's request and any explicitly approved plan define the authorized scope of work. - Choose the minimal solution sufficient for the task. Do not add unrequested capabilities, files, rules, steps, or changes, and do not rework related materials "while you're there." - Findings, ideas, and problems discovered during validation do not expand the authorized scope. - Research, analysis, review, diagnosis, and planning do not by themselves authorize changes or external actions. - Use update_plan for multi-step tasks (>1…

Reads credentials

We split this AGENTS.md into the rules it gives an agent

8 rules. About 1 is common, 6 are distinctive, and 1 is very specific to this project.

≈1
Common · estimate
Many files like this have a rule of this kind.
≈6
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

Distinctive:
NEVER use request_user_input tool.
line 8
Distinctive:
Never create Git worktrees or extra project copies unless the user explicitly asks.
line 9
Distinctive:
Use update_plan for multi-step tasks (>1 step).
line 22
Distinctive:
When the user asks for a team or swarm, orchestrate bounded subagent work with the runtime's team tools.
line 23
Common:
NEVER ask user to write secrets in chat
line 28
Distinctive:
Instead: provide instructions where to store them securely
line 29
Distinctive:
ALWAYS ask before pushing to main
line 32
Very specific:
ALWAYS add secrets to .gitignore: .env, *.key, credentials.json, secrets/
line 33

≈ “Common” is an estimate: at least 1 in 5 of the 36,368 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.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

AGENTS.md

Language

Artifacts addressed to the user (chat, plans, plan-mode, interviews, validator summaries, user-spec, README): the language the user writes in. Declare your language here — e.g. "This user writes in English." Skills and agents read the user's language from this line.
Technical docs, code, code comments, AI prompts, internal logs (tech-spec, tasks, AGENTS.md, skills): English.

Behavior

No "Great question!", no filler, no water.
NEVER use request_user_input tool. Ask questions as plain text in chat instead.
Never create Git worktrees or extra project copies unless the user explicitly asks.

Task Scope

The user's request and any explicitly approved plan define the authorized scope of work.
Choose the minimal solution sufficient for the task. Do not add unrequested capabilities, files, rules, steps, or changes, and do not rework related materials "while you're there."
Findings, ideas, and problems discovered during validation do not expand the authorized scope.
Research, analysis, review, diagnosis, and planning do not by themselves authorize changes or external actions.

Task Planning

Use update_plan for multi-step tasks (>1 step).
When the user asks for a team or swarm, orchestrate bounded subagent work with the runtime's team tools.

Security

NEVER ask user to write secrets in chat
Instead: provide instructions where to store them securely
- Local: .env files, config files
- CI/CD: GitHub Actions secrets
ALWAYS ask before pushing to main
ALWAYS add secrets to .gitignore: .env, *.key, credentials.json, secrets/
Be cautious with external actions (push, deploy, send messages, create PRs). Ask before acting externally when uncertain.

Download now (ZIP)

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 28,693 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 repository28 other files it gives its agents.
- skills/skill-master/references/agents.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- code-reviewingSkill
- code-writingSkill
- documentation-writingSkill
- infrastructure-setupSkill
- layout-reviewingSkill
- layout-writingSkill
- methodologySkill
- project-initializationSkill
- prompt-masterSkill
- security-auditorSkill
- skill-masterSkill
- test-masterSkill
- user-spec-planningSkill
- code-reviewingSkill
- code-writingSkill
- documentation-writingSkill
- infrastructure-setupSkill
- layout-reviewingSkill
- layout-writingSkill
- methodologySkill
- project-initializationSkill
- prompt-masterSkill
- security-auditorSkill
- skill-masterSkill
- test-masterSkill
- user-spec-planningSkill

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.

Check your setup, free