splicedeck
CLAUDE.md4 starsChanged 3 months agoApache-2.0View on GitHub
AGENTS.md above is the contract: twelve hard rules and the table that says which document owns which concern. It is imported, not summarised. A summary here would be a shortcut you could read instead of the source, and it would go stale first. This file carries only what AGENTS.md does not. Run the tests as python -m pytest. pyproject.toml sets testpaths, so the bare command finds everything. Do not copy a test path out of a document without checking it…
Files
@AGENTS.md
Claude Code
AGENTS.md above is the contract: twelve hard rules and the table that says which
document owns which concern. It is imported, not summarised. A summary here would
be a shortcut you could read instead of the source, and it would go stale first.
This file carries only what AGENTS.md does not.
Run the tests as python -m pytest. pyproject.toml sets testpaths, so the
bare command finds everything. Do not copy a test path out of a document without
checking it exists — several filenames cited in docs/ were never written.
Both surfaces are generated from splicedeck/surface/verbs.py. Edit that table,
never terminal.py or mcp.py's tool list. After any surface change run
python -m checks.golden --check, and regenerate with python -m checks.golden
when the diff is intended. The golden files exist so a surface change is reviewable.
The MCP server is spd serve, or python -m splicedeck.surface.mcp. Both reach
the same loop. --tools prints the generated tool list and exits, which is how you
tell a broken server from a broken host config.
A fixture must not fabricate its input. This is the expensive lesson in this
repository, not a style preference. A one-line stub named bookmarks/baseline.toml
kept the suite green for months while draft — and therefore every verb
downstream of it — could not run at all on a real clone. Media fixtures encoded
-preset ultrafast, which disables B-frames, so no test ever saw the cadence
misclassification that blocked every camera export. When a test needs shipped data,
read the shipped data.
Git has no commits yet. A 223 MB .mov sits at the repository root; it is
covered by .gitignore, and git check-ignore is how you confirm that rather than
assuming it. Do not stage or commit anything unless asked.
Where the open questions are. docs/roadmap.md Open items and the Open
sections of each owner document are genuinely undecided. AGENTS.md says not to
resolve one silently, and the ones most likely to tempt you are the twelve
selectable bookmarks (docs/bookmarks.md B-2) and the axes they are points in
(B-1). Inventing them would fix public names by accident.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/splicedeck.mdcCursor rule
- contributing-to-splicedeckSkill
- cutting-a-deliverableSkill
- cutting-vertical-clipsSkill
- recovering-from-a-refusalSkill
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
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.

