agentleFS
mariomilemariomile/builder-os/skills/agents-md/SKILL.md

agents-md

Skill2 starsChanged 49 days agoMITView on GitHub

Keeps the coding agent's environment working: writes or prunes AGENTS.md and CLAUDE.md, and runs retros that turn what slowed a session into pointers, checks or cuts. Use when a repo has no AGENTS.md, one is stale or long, or for a retro.

36 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use jira, how and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • MicrosoftSkill

    jira

    Jira issue workflows for search, issue updates, transitions, comments, field discovery, and interactive credential setup via the Jira REST API. Use when you need to configure Jira access, search with JQL, inspect an issue, create or update work items, move an issue between statuses, post comments, or discover required fields for issue creation.

    Also in 3 projects that use this

  • jaydubya818Skill

    how

    Use for runtime, ownership, layering, code-walkthrough, and placement questions.

    Also in 3 projects that use this

  • asterkinSkill

    adr

    Create Architecture Decision Records (ADRs) following project standards. Guides template usage, process, and best practices for documenting architecturally significant decisions.

    Also in 3 projects that use this

  • jaydubya818Skill

    why

    Use when design rationale must be recovered from source history and available records.

    Also in 3 projects that use this

  • photon-groveSkill

    unslop

    Detects and rewrites generic, overly polished, or AI-sounding prose while preserving meaning. Use when the user asks to "unslop", humanize, de-AI, tighten, or review whether writing sounds generated.

    Also in 4 projects that use this

  • AWSSkill

    wait-what

    Stop. That last message did not land: re-pitch it.

    Also in 3 projects that use this

The file

Files
SKILL.md

AGENTS.md

AGENTS.md (and CLAUDE.md, which imports it) loads into every session that opens the repository, so every line costs on every turn. Its job is to be an index: point at the material the agent cannot find in one search, give the commands it would otherwise guess, and hold the few rules that must always hold. Everything else lives where it is read on demand. Works on any repository, with or without BuilderOS. The retro mode, invoked by name (retro) or at the end of a build or an outcome review, reads a session and turns what slowed or misled the agent into changes to its environment: follow retro, with its own marker.

What goes where

MaterialHome
Where things are: entry points, the module holding the rules, where tests live, generated filesTECH.md → Map (or a docs/ file the index points at)
Build, test and lint commands an agent would guess wrongAGENTS.md, one line each; on a BuilderOS project TECH.md → Verify, pointed at
Rules that must hold every time (never edit generated files, never deploy without approval)AGENTS.md, each with the check that enforces it when one exists
A mechanical rule (a banned call, an import shape, a file location)A check: lint rule, hook, CI step. AGENTS.md names the check, not the rule
Conventions that need judgmentTECH.md → Conventions, read by the reviewer
Product, domain and decisionsPRODUCT.md, DOMAIN.md, decisions/, pointed at
Host-specific instructions (Claude Code only)CLAUDE.md below its @AGENTS.md import

A pointer is one line: what the material is, and when to read it, with the trigger word first ("Changing an export: read src/export/README.md"). The wording decides whether the agent reaches the material, so sharpen a weak pointer before inlining what it points at.

Procedure

  1. Read what exists. AGENTS.md, CLAUDE.md, README, the manifests and their scripts, the CI workflow, any pre-commit hook, docs/, and on a BuilderOS project TECH.md, DOMAIN.md and the BuilderOS block. Done when you can name the build, test and lint commands, the entry points and every rule already written down.
  2. Sort every existing line into the table above: keep, move (to which home), turn into a check, or cut. A line is cut when it states what the agent does by default, what one look at the repository answers, or what is no longer true (a path that does not exist, a command that fails). Done when every line has a verdict.
  3. Draft the index. Commands, pointers, must-hold rules, in that order, then the BuilderOS block on a BuilderOS project (references/builderos-state-schema.md, Session Start). Write each rule as the behavior wanted ("write migrations with the generator in db/"); a prohibition stays only as a hard guardrail, paired with what to do instead. Aim for well under 150 lines. Make CLAUDE.md @AGENTS.md plus Claude-only lines; create it when missing. Done when every kept or new line is a command, a pointer with its trigger word first, or a must-hold rule.
  4. Draft the homes. TECH.md → Map from the repository (on a BuilderOS project; elsewhere a docs/ file or the README section the index points at), and each moved line in its new place. Done when every line marked move in step 2 has a home and every pointer in the index lands on one.
  5. Show the diff, then write. Lines they wrote are moved or cut only with their yes, one round per pressure-testing for the moves and cuts that are not obviously dead; new lines and fixes to dead paths need no question. Done when the files are written and every path they name exists (scripts/bos.mjs gate 5 checks it as 5.11 on a BuilderOS project).

On a BuilderOS project this runs at initialization, at the end of each build (phase 5, step 11) and when a retro or the briefing flags a dead pointer or a file over 150 lines.

Capabilities

CapabilityUsed forFloor if absent
repo.readManifests, CI, entry points, existing steering filesAsk for the commands and the layout; mark the map unverified
files.writeThe index and its homesReturn the full files for the person to apply
shell.execRunning the commands once to confirm they workWrite them as found and say they were not run

Output Contract

## AGENTS.MD UPDATED with the line count before and after, what was moved and where, what became a check, what was cut and why, and the commands confirmed to run.

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.