register-roast
Skill2 starsChanged 22 days agoMITView on GitHub
Post-roast data closeout in one step — capture the operator's rating (the D42 label) and register the completed roast as a labelled replay fixture (store_to_fixture). Use after a roast to "register the roast", "rate and save the roast", or feed it to the corpus.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- capture-agent-usageSkill
- add-bean-profileSkill
- captureSkill
- pre-roast-preflightSkill
- pr-preflightSkill
- roast-reviewSkill
- triage-prSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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
- Very specific: line 14
Use that <RID> (a completed run).
- Distinctive: line 20
If the server exposes a rating endpoint, POST it.
- Very specific: line 30
Run it with the real <RID> + <slug> ($ROASTPILOT_DB defaults to ~/roasts/roastpilot.sqlite3).
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The closeout the roast-review skill prompts: turn a finished roast into a
rated, labelled corpus/eval example (D42 / D44). Run for the latest completed
run unless given an id.
1. Resolve the run
!curl -sf -m5 http://127.0.0.1:8000/api/roasts | python3 -c "import json,sys; r=sorted(json.load(sys.stdin).get('runs',[]),key=lambda x:x.get('started_at_utc',''),reverse=True); print(r[0]['id'], r[0].get('outcome'), r[0].get('started_at_utc','')[:19]) if r else 'no runs'"
Use that <RID> (a completed run). Pick a short <slug> for the bean (e.g. colombia-huila-roast5).
2. Rate it (the D42 label)
Capture the operator's score (e.g. 1–5) + notes (taste, what to change).
Persist it on roast_runs (operator_rating / operator_notes):
- If the server exposes a rating endpoint, POST it; otherwise have the operator set it in the UI (the RoastRating widget on the roast detail / history page) and confirm it saved. The rating is the corpus LABEL — do not skip it.
3. Register the fixture (D44)
Emit the labelled replay fixture (the format the bake-off / eval consume):
!echo 'python scripts/store_to_fixture.py "$ROASTPILOT_DB" --out-dir ~/roasts/fixtures/<slug> --run-id <RID>'
Run it with the real <RID> + <slug> ($ROASTPILOT_DB defaults to
~/roasts/roastpilot.sqlite3). It writes roast.jsonl + a labelled
summary.json carrying the degree classification (core_medium ≤195 /
soft_medium (195,197] / over >197) and the operator_rating / operator_notes.
4. Confirm
- The fixture parses and carries the label fields:
!
echo 'python -c "from scripts.bakeoff_replay import load_roast; e=load_roast(\"<out-dir>\"); print(e)" # spot-check it loads with the label' - Report the fixture path. It is gitignored — real stores / fixtures are never committed (privacy invariant); only aggregates + anonymised ids go in the repo.
- Note: validate store→fixture conversions on a REAL store, not synthetic (the telemetry/event clock mismatch silently reads the wrong row otherwise).
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.

