agentleFS
MesteriisMesteriis/Engineering-Bible-AI/AGENTS.md

Engineering-Bible-AI

AGENTS.md0 starsChanged 4 months agoMITView on GitHub

This repository packages portable engineering instructions, standards, skills, templates, validation, and installers. Treat it as a public distribution: local runtime state and private infrastructure do not belong in tracked files. Select the narrowest directly relevant skill. Do not invoke workflow-router for a clear single-domain request; reserve it for ambiguous or genuinely mixed work. Use one primary skill and at most one supporting skill by default. For a same-task follow-up in the same thread, reuse the current skill route and already-loaded…

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

6 rules. About 1 is common, 4 are distinctive, and 1 is very specific to this project.

  • ≈1

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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
  • could be automatic: a setting could enforce it
  1. Common:

    Do not commit credentials, auth files, provider endpoints, environment files, private keys, local agent sessions, caches, generated …

    line 39
  2. Distinctive:

    Prefer the smallest correct change.

    line 55
  3. Very specific:

    Use apply_patch for hand-authored edits.

    line 60
  4. Distinctive:

    Never claim a check passed unless the exact command ran successfully.

    line 62
  5. Distinctive:

    Use the smallest profile that proves the change, then run the broader profile when shared tooling, installation, or release behavior …

    line 66
  6. Distinctive:

    Before completion, review the diff for public behavior, runtime-boundary drift, generated documentation drift, missing regression coverage, …

    line 82

≈ “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

Engineering Bible AI Repository Instructions

Purpose

This repository packages portable engineering instructions, standards, skills, templates, validation, and installers. Treat it as a public distribution: local runtime state and private infrastructure do not belong in tracked files.

Routing Discipline

Select the narrowest directly relevant skill. Do not invoke workflow-router for a clear single-domain request; reserve it for ambiguous or genuinely mixed work. Use one primary skill and at most one supporting skill by default.

Continuation Fast Path

For a same-task follow-up in the same thread, reuse the current skill route and already-loaded instructions. Do not route again or reread an unchanged SKILL.md unless the domain, risk, required tools, or requested workflow changed. If compaction removed required instructions, reload only the missing skill.

Repository Sources Of Truth

  • skills/registry.yml owns skill membership, grouping, and order.
  • Generated registry blocks in README.md, README.ru.md, and MANIFEST.md must be updated with python3 scripts/registry.py --root . docs --write; do not edit their contents by hand.
  • engineering/README.md indexes the standards library. Add or remove a standard and its index entry together.
  • instructions/global/steady.md is the default installed global prompt.
  • instructions/global/full.md preserves exhaustive routing for strict use; instructions/global/minimal.md is the compact steady-state profile.
  • Root AGENTS.md contains project-only rules and must not duplicate either global prompt.

Runtime Boundary

Do not commit credentials, auth files, provider endpoints, environment files, private keys, local agent sessions, caches, generated capability catalogs, or machine-specific configuration. Runtime tools are discovered dynamically from the current session; tracked policies must describe capabilities and risk, not hard-code provider or tool identifiers.

Generated project-local capability files belong under .engineering-bible/ and must remain untracked. Do not claim a runtime capability is available until the current session exposes it.

Change Rules

  • Inspect relevant implementation, tests, docs, and current Git state before editing.
  • Preserve unrelated dirty-worktree changes; never reset, stash, or overwrite them without explicit authorization.
  • Prefer the smallest correct change. Keep installer, validation, registry, runtime discovery, and documentation responsibilities separate.
  • Public CLI changes require compatibility handling, tests, and matching docs.
  • Validation output must distinguish PASS, FAIL, and SKIP. A skipped release gate is not success.
  • Use apply_patch for hand-authored edits. Generated-block updates and formatting may use their deterministic project commands.
  • Never claim a check passed unless the exact command ran successfully.

Validation

Use the smallest profile that proves the change, then run the broader profile when shared tooling, installation, or release behavior changes:

make validate-quick
make validate-bootstrap
make validate
make validate-release

Tests are discovered with:

python3 -m unittest discover -s tests -p 'test_*.py' -v

Before completion, review the diff for public behavior, runtime-boundary drift, generated documentation drift, missing regression coverage, and accidental changes outside scope. Report exact commands, outcomes, assumptions, and remaining risks.

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.