agentleFS
Sign inStart free
Asaf-DahanAsaf-Dahan/super-skill/AGENTS.md

super-skill

AGENTS.md1 starsChanged 7 months agoMITView on GitHub

This is a Super Skill repository. It contains structured domain knowledge organized in 8 layers. Your role is to read, apply, and maintain this knowledge on behalf of the owner. This file is self-contained. You do not need to read another file to understand the operating rules. Everything you need is here. Note: This file is the entry point for non-Claude agents. It uses the lightweight SUMMARY.md-first protocol. Claude Code uses SKILL.md or CLAUDE.md which follow a different load sequence…

  • 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 repository5 other files it gives its agents.

The file

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

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    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
  1. Distinctive:

    Read files in this order before any task:

    line 21
  2. Distinctive:

    Never load a file the task does not require.

    line 42
  3. Distinctive:

    Write all proposals to PENDING.md.

    line 60
  4. Distinctive:

    Do not modify any layer file without explicit user approval.

    line 61
  5. Distinctive:

    If multiple agents are writing to this repository:

    line 79
  6. Distinctive:

    Before writing a new PENDING item, check PENDING.md to avoid duplicates

    line 83
  7. Distinctive:

    Use unique proposal IDs: EVAL-NNN, DRIFT-NNN, UPDATE-NNN

    line 84

≈ “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.

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

AGENTS.md

Super Skill - Instructions for Any AI Agent

<!-- Replace this header with your own project/organization name -->

What This Repository Is

This is a Super Skill repository. It contains structured domain knowledge organized in 8 layers. Your role is to read, apply, and maintain this knowledge on behalf of the owner.

This file is self-contained. You do not need to read another file to understand the operating rules. Everything you need is here.

Load Protocol

Note: This file is the entry point for non-Claude agents. It uses the lightweight SUMMARY.md-first protocol. Claude Code uses SKILL.md or CLAUDE.md which follow a different load sequence designed for its specific capabilities.

Read files in this order before any task:

  1. SUMMARY.md -- 80-line overview of the full domain (read this first, always)
  2. PENDING.md -- what is waiting for user approval

Load additional files only if the specific task requires them:

If the task involvesLoad this file
Architecture or tool decisionDECISIONS.md
Tool version or current statusCURRENT_STATE.md
Domain structure or sub-domainsDOMAIN_MAP.md
Evaluating something newEVALUATION.md
Generating learning contentLEARNING.md
Expert council or debatesexperts/COUNCIL.md
Execution trace or causal analysistraces/
Approval actionPENDING.md

Individual expert profiles (experts/[name].md) load on demand only, not by default. Load them when a specific expert's perspective is needed.

Never load a file the task does not require.

Capability Map

File-only (no shell required):

  • Reading and explaining any layer file
  • Writing proposals to PENDING.md
  • Generating learning content to notebooks/

Shell required:

  • python scripts/update_summary.py -- regenerate SUMMARY.md
  • python scripts/feed_notebook.py -- push files to NotebookLM
  • python scripts/generate_learning.py audio | quiz | mindmap

The One Rule

The model proposes. The user decides. The Super Skill records. The system executes. Write all proposals to PENDING.md. Do not modify any layer file without explicit user approval.

What Requires Approval

  • Any change to CURRENT_STATE.md
  • Any new entry in DECISIONS.md
  • Any adoption or rejection in EVALUATION.md
  • Any change to DOMAIN_MAP.md

What Does Not Require Approval

  • Reading and explaining existing content
  • Writing proposals to PENDING.md
  • Generating learning modules to notebooks/
  • Running scripts

Multi-Agent Use

If multiple agents are writing to this repository:

  • Each agent writes proposals to PENDING.md independently
  • The user resolves all conflicts and approves decisions
  • Before writing a new PENDING item, check PENDING.md to avoid duplicates
  • Use unique proposal IDs: EVAL-NNN, DRIFT-NNN, UPDATE-NNN

Slash Commands

For ongoing tasks, use slash commands: /ss-eval, /ss-learn, /ss-pending, /ss-sync, /ss-summary, /ss-drift, /ss-council, /ss-expert, /ss-synthesize

Using with Gemini

Gemini can operate a Super Skill the same way Claude Code does, with one adjustment: Gemini does not auto-load files on session start, so you load the entry points yourself.

Gemini CLI (gemini command-line tool): from inside the Super Skill folder, paste the contents of SUMMARY.md and SKILL.md into your first message. Then ask Gemini to read PENDING.md and the relevant layer file for the task. Gemini will follow the Iron Principle as long as you keep SKILL.md in context, because the rules are encoded there.

Gemini Code Assist (VS Code or JetBrains): open the Super Skill folder as a workspace. Use the chat panel and reference @SUMMARY.md @SKILL.md @PENDING.md in your prompt. Gemini Code Assist supports the @file syntax to attach repository files directly to the request.

Gemini in any other editor: same pattern as Cursor. Open the folder, ask the agent to read ONBOARDING.md, and run the prompt that matches your task. The Iron Principle applies regardless of which Gemini surface you use.

Compatible Agents

This Super Skill follows the Agent Skills open standard. It works with Claude Code, Cursor, Codex, Gemini, and any agent that can read and write markdown files.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free