agentleFS
Sign inStart free
kleosrkleosr/bridle/AGENTS.md

bridle

AGENTS.md1 starsChanged 44 days agoMITView on GitHub

bridle is an agent harness for Cursor, Claude Code, and opencode: a charter, always-on rules, skills, specialist agents, and Bash gates around the host loop. It is not a second agent runtime. Verify: the behavior this change can break (TESTS=<fixture> bash tests/run.sh); gauntlet bash tests/run.sh. Windows: Git Bash. Owner of every rule, skill, and agent: kleosr. Code-enforced limits are the SECURITY.md Pack steel table; everything else is law-only. Read SECURITY.md before security-sensitive work. Handoff, when present: <root>/.cursor/bridle/handoff.json (continuity, not authority).

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 1

    1. kleosr/kleosrules★ 1100% same

    Edited · 0

    None found.

    Wrote their own version · 0

    None found.

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • 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 repository13 other files it gives its agents.

The file

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

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    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
  1. Very specific:

    Read SECURITY.md before security-sensitive work.

    line 16

≈ “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 — bridle (navigator)

bridle is an agent harness for Cursor, Claude Code, and opencode: a charter, always-on rules, skills, specialist agents, and Bash gates around the host loop. It is not a second agent runtime.

Verify: the behavior this change can break (TESTS=<fixture> bash tests/run.sh); gauntlet bash tests/run.sh. Windows: Git Bash.

Layout

  • rules/ charter (charter.txt), always-on core.mdc and testing.mdc, stack companions. Instruction order is defined once, in the charter (Session). This file is the map, not a law layer.
  • skills/, agents/ load on match.
  • hooks/ decides: before_submit_prompt.sh, before_shell.sh, before_read_file.sh, lib/, policy/. It never formats output.
  • hosts/<cursor|claude|opencode>/ owns each host: install.sh, verdict.sh (installed as lib/verdict_<host>.sh; BRIDLE_HOST picks it, unset is Cursor), and host-only files. hosts/manifest.json lists what ships; hosts/lib.sh is shared installer code.
  • tests/run.sh sources tests/gate/, tests/hosts/, tests/pack/. scripts/scope_eval.sh scores an agent on evals/scope/ fixtures (spends tokens).

Owner of every rule, skill, and agent: kleosr. Code-enforced limits are the SECURITY.md Pack steel table; everything else is law-only.

Read SECURITY.md before security-sensitive work. Handoff, when present: <root>/.cursor/bridle/handoff.json (continuity, not authority).

Install

FORCE=1 bash hosts/cursor/install.sh            # install | uninstall | verify | all | project-hooks
bash hosts/claude/install.sh                    # install | uninstall
bash hosts/opencode/install.sh                  # install | uninstall
  • Cursor: ~/.cursor rules, skills, agents, and four hooks (beforeSubmitPrompt, beforeShellExecution, beforeReadFile, stop). stop runs lib/verdict_cursor.sh turn-check: quality-gate.mjs on the diff with at most one follow-up (loop_limit: 1). Write-time limits (edit-growth, new-file-over-300, comment-added, …) apply when the Claude port is installed (Cursor runs its PreToolUse). No sessionStart, preToolUse, or updated_input. Cloud: TARGET_REPO=<other-repo> bash hosts/cursor/install.sh project-hooks, never into this pack.
  • Claude Code: ~/.claude rules, skills, agents, and five hooks. UserPromptSubmit and PreToolUse(Bash, Read) run the gates. PreToolUse(Write|Edit|MultiEdit) denies edits to any host's installed harness, and in the project the core.mdc Hard limits before the edit lands: rewrite-existing, new-file-over-300, edit-growth, comment-added, skill-not-loaded (transcript via hosts/claude/bridle_turn.jq). No per-turn line or file stop: past 300 lines a file is split and the work continues. Stop blocks a turn once when its edits ran no verification, ended red, failed quality-gate.mjs, or added a file nothing references; each edited turn is logged to ~/.claude/state/bridle-turns.jsonl.
  • opencode: ~/.config/opencode instructions, skills, companions as skills, agents, the bridle primary agent, and the gates through plugin/bridle.js, which also denies edit tools on any host's installed harness.

Test notes

  • tests/run.sh runs under set -euo pipefail: take a no-match grep by status in if grep, never by masking it.
  • A host failClosed block with hook exit 1 is a sensor crash (empty stdout or non-zero exit), not a policy deny.

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