agentleFS
aalishan001aalishan001/opencockpit/.claude/skills/chassis/SKILL.md

chassis

Skill2 starsChanged 33 days agoMITView on GitHub

Example maintenance tracker for a vehicle (bike, scooter, car): owner's manual setup, service slip images, odometer readings, what's due, service history, maintenance intervals. Commands - configure-manual, configure-history, odom, deadlines, service, history, manual. Trigger even on bare inputs like "odo 15230" or a photo of a service slip.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 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.

  • 4

    Very specific

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

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

    All deadline math is done by python scripts/chassis.py.

    line 11
  2. Very specific:

    If a job has no km interval in the manual, find it on the web for this exact model and record source: web:<url>.

    line 20
  3. Very specific:

    Read every slip into a candidate row: date, odo, jobs (mapped to manual ids), workshop, source file.

    line 27
  4. Very specific:

    Ask for the current odometer reading, append it, write services.jsonl, commit, then run deadlines.

    line 31
  5. Distinctive:

    Never infer an odometer value from neighbouring slips.

    line 46

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CHASSIS

No command given (bare /chassis): read COMMANDS.md and print its ## CHASSIS section verbatim (heading through the line before the next ## ) in a fenced code block. Nothing else, no commit.

A worked example of a domain: maintenance tracking on kilometres (months-or-km "whichever earlier" → keep the km only, by design). Exception: a job the manual gives no km interval for (and none is found on the web) goes in a separate time-based category with interval_months. All deadline math is done by python scripts/chassis.py. Never compute due km yourself.

Data: data/chassis/manual.json, services.jsonl, odometer.jsonl, anomalies.jsonl (schemas in scripts/schemas.md).

One-off setup

configure-manual

Input: owner's manual (PDF/images) in inbox/ (or the optional second inbox set in COCKPIT_DRIVE_INBOX). Manual page images go through python scripts/ingest_images.py chassis <paths> first; PDFs are moved as-is.

  1. Extract every maintenance job: id, name, action, interval_km, and a source page reference.
  2. If a job has no km interval in the manual, find it on the web for this exact model and record source: web:<url>. If the web has nothing model-specific, store it with interval_months from the manual (time-based category), never a guessed km. Fully itemized: one row per item per distinct action/interval (e.g. brake_fluid replace 21000 and brake_fluid_check check 3000). Keep the manual's code (I, R, I & T, …) in notes.
  3. Show the full table, write manual.json, move originals to raw/chassis/, commit.

configure-history

Input: bulk service slip images in inbox/ or the second inbox. 0. python scripts/ingest_images.py chassis <paths> (or --pending to find them). Work from the printed raw/chassis/ copies; each row's source is that file name.

  1. Read every slip into a candidate row: date, odo, jobs (mapped to manual ids), workshop, source file. Keep unmapped job lines in notes.
  2. Unreadable or ambiguous fields go to a question list, not into data.
  3. Sort by date and run python scripts/chassis.py check-candidates <tmpfile> to detect anomalies: decreasing odo, implausible jumps, date/odo order conflicts.
  4. Present the timeline plus each anomaly with a proposed resolution. I decide. Log decisions to anomalies.jsonl.
  5. Ask for the current odometer reading, append it, write services.jsonl, commit, then run deadlines.

Commands

CommandTriggerDoesEnd
odom <km>command + readingpython scripts/chassis.py odom <km> (appends, rejects decreases, flags jumps), then shows deadlinesone-off
deadlinescommandpython scripts/chassis.py deadlines: every job with last done (odo + date) and km remaining, sorted soonest first; overdue at topone-off
servicecommand + imageIngest the image (ingest_images.py chassis <path>; source = the raw/ copy) → extract slip → show parsed row → ask only about unclear items → python scripts/chassis.py add-service <json> → show deadlines → commitone-off (asks if needed)
historycommandpython scripts/chassis.py historyone-off
manualcommandpython scripts/chassis.py manualone-off

Extraction rules for slips

  • Workshop shorthand varies. Map to manual ids by meaning (e.g. "EO chg" → engine_oil). If unsure, ask.
  • "Checked" vs "replaced" matters: only count a job as done if the slip shows it was actually performed.
  • Never infer an odometer value from neighbouring slips. Missing odo → question.

Download now (ZIP)

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.