lego-check

Checks a lego model: all studs connected or stubbed, no duplicate providers, slots filled, units and frequencies consistent, identities verifiable, simultaneous blocks listed.

SKILL.md

GitHub
<!-- Generated from .claude/skills/lego-check/SKILL.md by tools/sync_agents.py. Edit the source, then run the script. -->

/lego-check — Lego Check

Category: Lego (Modular Models) · Primary agent: model-validation-specialist

Checks a lego model: all studs connected or stubbed, no duplicate providers, slots filled, units and frequencies consistent, identities verifiable, simultaneous blocks listed.

Arguments: whatever the user wrote after the command name.

Purpose

Check that a lego model is wired correctly and that the manifest tells the truth about the code: every stud connected or stubbed, no duplicate providers, every slot filled once, units and frequencies consistent, timing as declared, identities holding numerically, and simultaneous blocks listed. The output is a verdict that blocks release while bricks are unvalidated or stale.

Use it when after any change to a lego model, before a forecast round or release, or when results look inconsistent.

Do not use it when the brick-level economic validation is needed (/validate-model).

Before you start

  1. Check required inputs. For each input below record FOUND or ABSENT (never 'assumed'):
    • A model.lego.json manifest
    • The model code or specification (to compare the declared ports with what is implemented) If a required input is ABSENT, that part is NOT ASSESSED — NO DATA: do not estimate or invent it; name what is missing and which skill produces it. If all required inputs are absent, stop and say so.
  2. Check confidentiality. Identify the data classification level of the inputs (.claude/docs/confidentiality-and-ai-use.md). Do not proceed with restricted or confidential data in a way the policy does not allow.
  3. Collaboration. Follow .claude/docs/collaboration-protocol.md: ask questions first, present options, get approval before writing files, and update production/session-state/active.md after each approved section.

Agents

  • model-validation-specialist — primary agent; coordinates the work and integrates results
  • modeling-methodology-specialist — Chooses methods and architecture across model families, sets code and data standards, and resolves technical trade-offs between traditional, structural and AI approaches
  • economic-reasoning-specialist — Frames the economic question and tests whether results make economic sense

Each agent is defined in .claude/agents/<name>.md. If your tool runs subagents, run these agents as subagents, in parallel where the steps are independent; otherwise read each agent's file and apply its instructions in turn. Subagents only write to the paths named in your instructions.

Workflow

  1. Run python tools/lego.py check <manifest> and report errors, warnings and notes
  2. Compare the manifest with the code: every declared provided variable exists in the code with that unit; no variable crosses a brick boundary without being declared
  3. Check timing: lagged joints are really lagged in the code; same-period joints sit in a simultaneous block that the solver handles jointly
  4. Check frequencies: each brick on a different frequency has a declared bridge, and the bridge is implemented and tested
  5. Check identities: each declared accounting identity holds numerically in a run, not only on paper
    • Recount each brick's elements from its code (python tools/lego.py elements <manifest> <brick> --from-mod <file>) and compare with the declared counts; explain any non-square brick and mark constraints verified only after a run confirms them
  6. List bricks that are unvalidated or stale and block release until they are validated

Checkpoints with the user

  • After the tool output: review errors before the manifest-to-code comparison

Common pitfalls

  • Trusting the manifest without reading the code
  • Undeclared variables passed between bricks through global state
  • Identities checked symbolically but not in a run
  • Bridges declared but not tested

Output

(conversation); optionally validation/<model>/lego-check.md

The check report lists tool errors and warnings, manifest-to-code findings, timing and frequency results, identity residuals and brick validation status.

Gate / verdict

  • Verdict: PASS / PASS WITH CONDITIONS / FAIL / NOT ASSESSED (if the code is absent, the manifest-to-code comparison is NOT ASSESSED)
  • Identity residuals from an actual run are reported
  • Declared element counts match the code, or the difference is explained

Suggested next steps

  • /validate-model
  • /model-release