filmcraft/AGENTS.md

FilmCraft — rules for agents and contributors

These rules apply to every human and AI contributor. CLAUDE.md holds the working instructions; this file holds the rules that must never be broken. When the two disagree, this file wins.

0. Never crash

People trust FilmCraft with hours of work. A malformed file, a bad command or MCP parameter, a corrupt preset or project, an odd keystroke or a full disk must give an error the user (or agent) can act on, never a crash and never lost work. This rule outranks feature work: don't ship a feature by adding a panic path, and fix a crash before building on top of it. The cross-app standard is craftrules/standards/never-crash.md.

1. Fail with Result<T, E>. Return the crate's error type and propagate with ?; add context (map_err, an error variant) instead of discarding it. In the UI, report the error (status bar, error dialog) and carry on.
2. No panicking shortcuts outside tests: no unwrap(), expect(), panic!, unreachable!, todo! or unimplemented!. Use ?, ok_or(..)?, let Some(x) = .. else { return Err(..) }, if let, or a fallback that is truly correct (never one that silently corrupts a project). An unfinished feature returns an "unsupported" error. The only exception is a call that is provably infallible from local code alone (a literal that always parses): a single-item #[allow(clippy::expect_used)] with the reason.
3. No unsafe (unsafe_code = "forbid" for the workspace). The single exception is crates/platform, which needs unsafe to call the operating system's media APIs (hardware video decoding through VideoToolbox, Media Foundation / Direct3D 11, VA-API and NVDEC; hardware encoding) and does nothing else. It uses unsafe_code = "deny" with #[allow(unsafe_code)] only on its FFI modules, a // SAFETY: comment on every unsafe block, a safe Result-returning public API, and the pure-Rust decoder as the tested fallback (ADR 0001).
4. Every input-derived number is hostile. Media files, project and interchange files, presets, fonts, CLI / MCP / control-channel parameters and UI state are untrusted. Index and slice with get() (or validate bounds once, up front, for a hot loop); slice strings only at char boundaries; use checked_* / saturating_* for lengths, offsets and counts; never divide by a value that can be zero (frame rates, timescales, sample rates, sizes); don't cast negative or NaN values to integers; don't call clamp with bounds that can cross; cap allocations sized by input.
5. Bound recursion and loops. Projects can be cyclic (nested sequences) or deeply nested: walk them with seen-sets or depth limits.
6. Don't cascade. Lock poisoning is not fatal: lock().unwrap_or_else(PoisonError::into_inner). Every thread or job (export, previews, proxies, decoding workers…) runs under catch_unwind and reports a failure as an error: a dead worker leaves monitors blank or jobs "running" forever.
7. Last-resort guard. The panic hook (crates/ui-egui/src/crash.rs) logs every panic, and the UI pass runs under catch_unwind, so an escaped panic becomes an error window and the session (and unsaved project) survives. It is a safety net, not a substitute for rules 1–6. Keep panic = "unwind".
8. Prove it. Every crash fix comes with a small synthetic regression test that panicked before the fix. Parsers and decoders get mutation-fuzz tests (truncation, bit flips, corrupt sizes) run under catch_unwind; new commands get hostile-parameter tests. Fuzzer findings become tests.
9. Enforced: clean crates carry #![deny(clippy::unwrap_used, clippy::expect_used, clippy::panic, clippy::unimplemented, clippy::todo, clippy::unreachable)] (moving to [workspace.lints.clippy] once every crate is clean). clippy.toml allows them in tests: a failing test should fail loudly.

1. Assets: no Adobe artwork, every asset licensed and attributed

This rule is absolute. Breaking it is the most serious mistake a contributor can make on this project.

What counts as an asset: any image, icon, cursor, logo, illustration, screenshot, font, colour LUT, preset, template, sound, music, video, 3D model, or other non-code media. This covers files in the repository, bytes embedded in code (include_bytes!, base64, data URIs), and data hard-coded to reproduce an image (for example, point lists traced from someone else's icon).

1. Never use Adobe iconography, images or other Adobe assets, in any form:
   • no icons, cursors, logos, splash screens, UI artwork or screenshots from any Adobe product;
   • no Adobe fonts, LUTs, presets, templates, sound effects, stock media or sample projects;
   • no traced, redrawn, recoloured or "inspired-by" copies of Adobe icons. Our icons may follow generic conventions (a play triangle, a razor blade, a stopwatch), but each must be drawn from scratch without reference to Adobe's artwork.
2. Every asset must be open: open-source licensed (MIT, Apache-2.0, BSD, ISC, zlib, SIL OFL…), public domain / CC0, or Creative Commons (CC BY or CC BY-SA; no NC/ND licences), or created by a contributor who owns it and licenses it to the project under the project licence (MIT OR Apache-2.0). If the licence is unknown or unclear, the asset does not go in.
3. Every asset needs an attribution sidecar. Next to each asset file X, add X.attribution with:

   asset:        <file name>
   title:        <what it is>
   author:       <creator / copyright holder>
   source:       <URL, or "original work" for contributor-made assets>
   license:      <SPDX id or licence name>
   license-file: <path to the licence text, if the licence requires shipping it>
   added:        <YYYY-MM-DD> by <contributor>
   notes:        <how it was made or modified; for screenshots, what is shown>
   

   Also add one line for the asset to ATTRIBUTION.md. cargo xtask assets (part of cargo xtask ci) fails if any asset lacks a sidecar or index entry.
4. Screenshots in the repo may show only FilmCraft (or other open projects), with media we generated or media that is itself openly licensed. Never commit screenshots of Adobe products.
5. Local reference material stays local. Premiere reference screenshots and notes live only in plan/premiere/ (gitignored). They must never be committed, bundled, embedded, traced or shipped.
6. Code-drawn assets (icons in crates/ui-egui/src/icons.rs, procedural demo footage in crates/media, procedural looks in crates/render) are original work under the project licence and are listed in ATTRIBUTION.md.
7. When in doubt, leave it out and draw or generate it yourself.
8. The one exception: first-party ArtCraft brand marks. The ArtCraft name and logos in docs/brand/ are trademarks of the ArtCraft Team, not open source, usable only unmodified and only in the context of FilmCraft under docs/brand/LICENSE-brand.txt; forks and modified versions must remove them. They still need a sidecar and an ATTRIBUTION.md row (licence LicenseRef-ArtCraft-Trademark). No other non-open asset is allowed, and this exception never covers third-party marks (Adobe, Discord, GitHub and other logos stay out; draw a generic icon instead).
9. Fonts live in storytold/craft-fonts, never in this repo. Don't commit new font files (the small Latin UI fonts already in assets/fonts/ stay). A font FilmCraft needs is added to craft-fonts, which the app reads through the optional build input CRAFT_FONTS_DIR=<absolute path of a craft-fonts checkout> (crates/text/build.rs embeds the fonts in its fonts/manifest.txt as filmcraft_text::fonts::CRAFT_FONTS; unset, it is empty and the app uses its bundled and system fonts). Never add craft-fonts to a Cargo.toml. Code using CRAFT_FONTS must work when it is empty. Standard: craftrules/standards/fonts.md.

2. Clean-room code

• Never read, disassemble or copy anything inside Adobe application bundles; file names and listings only. Observe behaviour by using the app; never capture its Home screen, recent projects, account info or file browsers.
• Never copy GPL/LGPL/AGPL code (FFmpeg, x264, x265, MLT, Kdenlive, Shotcut, Olive, LAME…).
• Implement codecs and formats from public specifications (ITU-T/ISO/IEC standards, IETF RFCs, the VP9 and AV1 bitstream specs, SMPTE documents, published container specs). Do not read the source of reference or third-party decoders/encoders while implementing a format, even permissively licensed ones (libvpx, libopus, libaom, dav1d, openh264…); spec text and conformance vectors only. Record the spec edition used in the crate README.
• ffmpeg/ffprobe may be used only as external test oracles and fixture generators. They are never linked, bundled or shipped.
• Dependencies must use permissive licences: MIT/Apache-2.0/BSD/ISC/Zlib/Unicode/CC0/BSL-1.0, or MPL-2.0 used unmodified.

3. Engineering rules

3.1 Never crash

See §0. It outranks every other engineering rule.

3.2 The rest

See CLAUDE.md: pure Rust, dependency layering (cargo xtask layers), exact Tick time, everything is a command, everything is agent-drivable, and the quality gates (cargo xtask ci) before every commit.

Shared real-file test corpora (Photoshop-authored PSDs, etc.) live in storytold/photocraft-corpus, explained in craftrules standards/test-corpora.md. Never commit large binary fixtures to this repo; fetch them pinned by commit and sha256-verified, as PhotoCraft does with cargo xtask corpus.

See also

• CONTRIBUTING.md and docs/contributing.md: setup, gates, commits, how to add things
• docs/architecture.md: layers, data model, commands, pipeline
• docs/testing.md: oracle tests, criteria, benchmarks
• docs/agents.md: driving FilmCraft over MCP / the control channel, and the agent work loop
• docs/control-protocol.md: control-channel method reference
• ATTRIBUTION.md: asset index
• ROADMAP.md, docs/target-app-parity.md, docs/gaps.md: stage, parity with Premiere Pro and the ranked work list (craftrules standards/progress-docs.md)

Contributor credits (About window)

• About ▸ Contributors/Models are compiled into the binary from contributors/contributors.json (commit stats; generated, never hand-edit) and contributors/people.toml (names people chose for themselves). See docs/contributors.md.
• Agents working for a contributor: when you prepare a PR, check whether your human's GitHub username has a [people.<username>] entry in contributors/people.toml. If not, ask them once whether they want to be credited by more than their username: a real name, a display name, and/or their public GitHub profile name (sync_github_name = true). If yes, add only their own entry (copy the template at the top of the file, or run python3 ../../craftrules/scripts/contributors.py --add-me . --real-name "…" --sync-github-name) and include it in their PR, committed as them. If no, change nothing: they are credited as @username anyway.
• Never add, edit, guess or copy anyone else's entry or name (not from git config, commit authors or GitHub profiles). Never hand-edit contributors.json.
• Maintainers refresh the stats with python3 ../../craftrules/scripts/contributors.py . (it also re-verifies who wrote each people.toml entry; --check only verifies).