pre-roast-preflight
Skill2 starsChanged 22 days agoMITView on GitHub
Hardware + software readiness checks before charging beans — Hottop serial present, FC mic present, real driver (not mock), advisor reachable, the intended bean profile loaded, plus the cold-machine safety checks. Use before a supervised roast, after connecting hardware.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- capture-agent-usageSkill
- add-bean-profileSkill
- captureSkill
- pr-preflightSkill
- register-roastSkill
- roast-reviewSkill
- triage-prSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
If the serial device is absent: power on the Hottop, re-seat the USB *data* cable, re-check.
- Distinctive: line 27
Read the startup banner: real Hottop (not mock), FC mode audio + model loaded, advisor REACHABLE, the trace DB path.
- Distinctive: line 36
If any fails: stop and fix — do not roast.
≈ “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
Codify the readiness checks that caught a powered-off Hottop before roast 4. Produce a go / no-go with each check ✅/⚠️ and the single fix for any ⚠️.
1. Pre-launch (read-only, before roast-live.sh)
- Hottop serial present — must match the MCP config
port: !ls /dev/cu.usbserial-* 2>/dev/null || echo "⚠️ NO serial device — power on the Hottop / re-seat USB" - MCP config sane — real driver (NOT
mock), FC modeaudio, known-good model revision, the expected mic substring: !grep -nE "driver:|port:|mode:|input_device:|revision:" "${COFFEE_ROASTER_MCP_CONFIG:-$HOME/roasts/coffee-roaster-mcp.yaml}" 2>/dev/null || echo "⚠️ MCP config not found" - FC microphone present — the configured
input_deviceenumerates: !system_profiler SPAudioDataType 2>/dev/null | grep -iE "USB|PnP|input" | head || echo "check the mic" - Branch/seed — the intended bean profile + targets are on the checked-out branch (
git branch --show-current;grep target_ src/roastpilot_agent/seed.py).
If the serial device is absent: power on the Hottop, re-seat the USB data cable, re-check. Do not launch until it appears.
2. Once roast-live.sh is up
- Health —
GET /api/health:mcp_child: running, advisorreachable(NOT "unreachable" — the expired-key trap that killed attempt 1), real driver: !curl -sf -m5 http://127.0.0.1:8000/api/health | python3 -m json.tool 2>/dev/null || echo "server not up yet" - Intended profile loaded with the expected drop/dev targets:
!
curl -sf -m5 http://127.0.0.1:8000/api/bean-profiles 2>/dev/null | python3 -c "import json,sys; [print(p['name'],'drop',p['target_drop_temp_c'],'dev',p['target_development_percent']) for p in json.load(sys.stdin).get('profiles',[])]" 2>/dev/null || echo "server not up yet" - Read the startup banner: real Hottop (not mock), FC mode audio + model loaded, advisor REACHABLE, the trace DB path.
3. Cold-machine safety checks — BEFORE any beans
Surface these for the operator to perform (dry, cold machine). They are the gate:
- Emergency stop → heat goes to 0, a fault is recorded, the dashboard shows e-stop.
- Over-limit → a heat above the limit is CLAMPed / REJECTed in the trace (architecturally enforced; observe a CLAMP/REJECT naturally if no inject path).
- Restart → recovery → kill the agent mid-(dry)-run → on restart it enters
operator_recovery_requiredand commands nothing until the operator acts.
If any fails: stop and fix — do not roast.
4. Verdict
Report a clear GO (all ✅) or NO-GO with the one blocking ⚠️ and its fix. Remind: fire extinguisher in reach, ventilation, you stay at the machine, hand near the kill.
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.

