agentleFS
BaratiLabBaratiLab/mechanical-engineering-skills/skills/mechanical-teardown/SKILL.md

mechanical-teardown

Skill16 starsChanged 16 days agoMITView on GitHub

Reverse-engineers a physical product (power tool, appliance, consumer device, mechanism, vehicle part) as an engineering system — function structure, costed BOM, materials and processes, Boothroyd DFA index, should-cost, repairability, FMEA, design decisions worth copying, benchmark and redesign opportunities — and renders a print-ready HTML teardown report. Use when the user says "tear down", "teardown", "reverse engineer", "how is this made", "what would this cost to make", "why is this designed this way", or shares photos of a disassembled product or an exploded view.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

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

    Follow references/teardown_protocol.md for what to photograph and measure at each layer.

    line 24
  2. Very specific:

    Function structure and requirements: Write 4–8 user needs as measurable engineering metrics, and say how the design meets each one.

    line 26
  3. Very specific:

    Fill the spec: Run python scripts/teardown.py template > spec.json.

    line 29
  4. Very specific:

    Compute: Run python scripts/teardown.py analyze spec.json.

    line 40
  5. Distinctive:

    Elimination candidates: Check every part that fails the minimum-part test.

    line 43
  6. Distinctive:

    Assumptions: Every default the calculator applied is listed.

    line 46
  7. Very specific:

    Report: Run python scripts/teardown.py report spec.json -o <product>_teardown.html.

    line 56
  8. Distinctive:

    Numbers with provenance: Every mass and cost is measured, quoted, or estimated, and the report says which.

    line 63
  9. Distinctive:

    Name the trade-off: Every design decision costs something, such as repairability, noise, tooling or tolerance.

    line 65

≈ “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.

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

Mechanical Teardown

A teardown explains why a product is built the way it is, what it costs to make, and what the next design should change. Treat the product as a system. Start from what it must do. Use every part to explain a decision, and back every number with an observation or a stated estimate.

Workflow

  1. Frame it. Pin down the product (make, model, price, year) and why the user wants the teardown: competitive benchmark, cost-down, design for repair, learning, or patent and design-around work. The purpose decides which sections get the most depth. Note the annual volume. If it is unknown, assume one and say so, because tooling amortisation depends on it.
  2. Gather evidence, in this order of trust.

    1. The user's own disassembly: photos of each layer, kitchen-scale masses, caliper readings.
    2. Exploded parts diagrams and service manuals, which give part names and counts.
    3. Regulatory internal photos (FCC ID for anything with a radio) and patents (claims explain intent).
    4. Published teardowns and repair guides.
    5. Your own engineering judgement, labelled as an estimate.

    Follow references/teardown_protocol.md for what to photograph and measure at each layer. Never invent part counts you cannot support. When a section is hidden, say "not inspected".

  3. Function structure and requirements. Write 4–8 user needs as measurable engineering metrics, and say how the design meets each one. Then build a function tree (store/convert/transmit/control/ protect...) and map every part to a function. A part that maps to no function is a finding.
  4. Fill the spec. Run python scripts/teardown.py template > spec.json. For each part record id, name, qty, subassembly, kind (part / fastener / label / process, e.g. lubrication or regulation), material, process, mass_g, and either cost_usd (quote, catalogue price, or a purchased-part estimate) or the process inputs (cycle_s, cavities, tooling_usd). Then record the DFA fields:

    • handling: easy | orient | small | flexible | heavy
    • insertion: drop_in | align | snap | hold_down | press | screw | rivet | connector | solder | adhesive | weld | wire_route
    • the three Boothroyd questions: moves, different_material, separate_for_service

    Mark the parts that wear out or fail as wear_part, with depth (how many parts must come off first). If a step takes a known time (regulation, a measured operation), set time_s per item instead. Times and tables are in references/dfa_times.md; cost data is in references/should_cost.md.

  5. Compute. Run python scripts/teardown.py analyze spec.json. Read the output critically:
    • DFA index. Under 10 % is poor, 20–30 % is typical of good consumer products, and over 40 % is excellent.
    • Elimination candidates. Check every part that fails the minimum-part test. Each one is a possible consolidation, unless a reason outside the test applies (cost, tolerance, supply).
    • Cost Pareto. A handful of parts usually make 80 % of the cost; focus the cost discussion there.
    • Assumptions. Every default the calculator applied is listed. Replace the ones that matter.
  6. Judge. Fill these sections of the spec:
    • design_signals: 3–6 specific decisions worth copying, each with evidence and the trade-off it makes. This is the most valuable section, so make it concrete ("clutch ramps molded into the ring-gear face", not "clever design").
    • fmea: S/O/D on 1–10, with an action for every High-priority row.
    • benchmark: 2–4 real or typical rivals on 4–6 measurable dimensions.
    • opportunities: each a specific change, its impact, its effort, and the saving per unit (negative if it adds cost).
    • verdict: 3–4 sentences covering where it wins, where it leaks, and what the next design should do.
  7. Report. Run python scripts/teardown.py report spec.json -o <product>_teardown.html. Add teardown photos under images; they are embedded in the file, so it stays one self-contained page that prints cleanly to PDF. Publish it as an artifact or send the file. In chat, give the five headline numbers and the verdict. Do not repeat the report.

Principles

  • Function first, parts second. List parts only to explain how a function is delivered.
  • Numbers with provenance. Every mass and cost is measured, quoted, or estimated, and the report says which. Should-cost is for ranking and comparison, not a quote.
  • Name the trade-off. Every design decision costs something, such as repairability, noise, tooling or tolerance. Say what was paid for it.
  • Specific opportunities only. "Replace 4 solder joints with a keyed connector: −30 s assembly, motor swap in 2 min" is an opportunity. "Improve repairability" is not.
  • Respect IP and safety. Reverse engineering for analysis is fine. Do not help copy trade dress or bypass safety interlocks. Warn before opening capacitors, lithium packs, mains-powered devices, or springs under preload.

Related skills

dfm-review (rule-level DFM on one part), design-review-package (full FMEA and BOM for your own design), material-selection (alternatives for a part the teardown flags), machine-element-sizing (check a gear, spring or bearing you measured), tolerance-stackup (a fit you suspect is marginal).

Examples

  • examples/cordless_drill.json: a generic 18 V drill/driver (clamshell, planetary gearbox, slip clutch). Report.
  • examples/automatic_watch.json: a generic automatic diver's watch with the movement torn down (gear train checked by calculation, pawl-lever winding, lever escapement, keyless works, process steps such as lubrication and regulation with time_s). Its exploded view and gear-train figures are drawn by examples/make_watch_figures.py and embedded through images. Report.

Both are illustrative: the structure and mechanics are right, and the masses and costs are labelled estimates. Run the tests with python tests/test_teardown.py.

Credits

The idea of a fixed-spine teardown that ends in a printable HTML report comes from yanliudesign/product-teardown-skill, which does product-strategy teardowns of software. This skill is an independent mechanical-engineering counterpart, with its own method, calculator and report.

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.