agentleFS
skulitomskulitom/Cutbolt/AGENTS.md

Cutbolt

AGENTS.md1 starsChanged 22 days agoMITView on GitHub

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

9 rules. About 0 are common, 7 are distinctive, and 2 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.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Keep docs/PROGRESS.md current through tools/verify.py --thorough and the evidence in progress/capabilities.json.

    line 8
  2. Distinctive:

    Use CLI/library interfaces and optional MCP over stdio.

    line 11
  3. Distinctive:

    Do not add third-party application binaries, installers, plugins, assets, presets, documentation copies, SDK source, project files, memory …

    line 12
  4. Distinctive:

    Do not translate, paraphrase, or port decompiled implementation into the engine.

    line 13
  5. Very specific:

    Read docs/RESEARCH.md before any proprietary application inspection.

    line 14
  6. Distinctive:

    Do not claim a formal clean-room process merely because files or agent conversations are separate.

    line 15
  7. Distinctive:

    Do not vendor third-party code or binaries.

    line 16
  8. Distinctive:

    Never claim compatibility across all builds of a third-party application.

    line 19
  9. Distinctive:

    Keep product-specific research and audit records private and excluded from Git.

    line 20

≈ “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

Project instructions

  • Build an original video editing engine for local agent use. The user explicitly does not want to host an API.
  • This repository has a working Rust/JSON snapshot engine, transactional local sessions with retries/undo/history, a narrow FFV1/PCM renderer, a persisted Windows render queue, and native MCP stdio. Bounded PNG/WAV scene compilation and timeline frame/range previews are also implemented; see docs/SCENES.md for strict limits. Read README.md, docs/USAGE.md and docs/AGENT_INTERFACE.md for implemented behavior; architecture documents include future work.
  • For PixelForge/Qwen/YouTube work, read docs/YOUTUBE_PIPELINE.md, docs/pipeline/CONTRACT.md and docs/pipeline/ACCEPTANCE.md first. These describe the agreed pilot and prepared model assets; docs/pipeline/RESULTS.md distinguishes the passing bounded P1/P2 fixture from planned production work. Keep companion tools, weights and generated media external; reuse the existing session contract.
  • For distribution, agent skills/plugins, the MCP default catalog, graphics authoring or positioning against HTML video tools, read docs/HYPERFRAMES_STRATEGY.md first. It is a dated strategy recommendation, not a scope or points change.
  • Original project code is MIT licensed. Preserve LICENSE and package metadata; dependencies retain their own licenses.
  • Keep docs/PROGRESS.md current through tools/verify.py --thorough and the evidence in progress/capabilities.json.
  • Verify in tiers: cargo test and tools/verify.py --only <fixtures> while working; commit, then python tools/ship.py to gate, push and verify the rest in the background (fix forward anything tools/verify.py --status reports); the full quick tools/verify.py for larger merges; and --thorough only for milestones and progress claims. A commit may carry stale evidence when its PROGRESS_HISTORY entry says so. See docs/DEVELOPMENT_LOOP.md. Only award points for the specific verified acceptance criteria. Record progress/scope changes in docs/PROGRESS_HISTORY.md. Do not inflate editing coverage with agent-only features or silently remove desirable but unimplemented features.
  • The user accepted these exclusions: accounts/activation, cloud services, stock marketplaces, telemetry, and GUI-only workflows. Underlying editing operations, offline AI-assisted editing, and project compatibility remain in scope.
  • Use CLI/library interfaces and optional MCP over stdio. Do not introduce a listening network service, remote dependency at runtime, or telemetry without a new product requirement.
  • Do not add third-party application binaries, installers, plugins, assets, presets, documentation copies, SDK source, project files, memory dumps, disassembly, or decompiled code to this repository.
  • Do not translate, paraphrase, or port decompiled implementation into the engine. Renaming code or keeping the original file out of Git does not establish independent authorship.
  • Read docs/RESEARCH.md before any proprietary application inspection. Raw research belongs outside this repository. Implementation must use original design, public specifications, and appropriately reviewed factual compatibility requirements.
  • Do not claim a formal clean-room process merely because files or agent conversations are separate. An implementer exposed to proprietary implementation is not an independent clean-room implementer.
  • Do not vendor third-party code or binaries. Keep dependencies external, record exact versions and licenses when selected, and use original synthetic fixture generators.
  • Preserve source media. Use explicit input/output roots and never overwrite an input as part of rendering.
  • Store timeline time as exact rational values, and reject unsupported editing semantics explicitly.
  • Never claim compatibility across all builds of a third-party application; record and test an exact build for any compatibility work.
  • Keep product-specific research and audit records private and excluded from Git. Published documentation uses project-owned capability names. Run the repository material check before committing; preserve all provenance restrictions above.

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 repository7 other files it gives its agents.

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.