schematic-erc-review

SkillAI & models

RAG-backed. Review a DipTrace schematic with bounded ERC, logical connectivity, metadata, and BOM checks before layout. Use when the user says “Review this DipTrace schematic before layout.”

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the schematic-erc-review skill

What this skill tells your AI

The instructions your AI receives, as published by fireostendere/mcp_diptrace in skills/schematic-erc-review/SKILL.md and read by ahel’s review.

Read runtime access before choosing between explicit-path MCP, a live bridge session, and native/headless CLI. Their availability is separate.

Schematic ERC review

RAG: engineering memory by defaultshared workflow. Derive an engineering checklist from the circuit-analysis/design corpus and DipTrace schematic/ERC courses, then inspect the actual circuit beyond automated ERC findings.

Produce a read-only disposition with explicit implemented, skipped, and unavailable checks. Use public tools/list for exact callable names and get_capabilities for document/configured feature availability.

Workflow

  1. Call diptrace_status, get_capabilities, and get_document_info; require a schematic source and freeze its SHA-256.
  2. Read get_schematic_model and get_connectivity_graph. Preserve sheet identity, hierarchy, explicit no-connect state, RefDes, pin numbers, and net names.
  3. Run run_erc, run_schematic_review, run_connectivity_check, and run_bom_review only when advertised.
  4. For native opening/roundtrip evidence, use the headless schematic workflow on a protected copy through evidence capture. Generic native schematic ERC is not implemented by the base helper; report that specific missing check rather than saying headless Schematic is unavailable. Deduplicate by check ID plus object identity.
  5. Report every missing mandatory category in skipped_checks; a skipped mandatory check forbids completed.
  6. Validate the report with ../shared/result.schema.json.

Quantitative boundaries

  • Query and report pages default to 100 records; follow returned continuation metadata rather than assuming one page is complete.
  • Confidence is a number from 0 through 1 in the shared schema. It is not a substitute for an evidence class.
  • SHA-256 identity is 64 lowercase hexadecimal characters and must remain unchanged throughout the read-only run.
  • All geometric distances are millimetres, independent of document Units.

The implemented review profiles live in review.py, advanced_review.py, and are summarized in REVIEW_ENGINE.md. Do not claim checks outside that matrix.

Refusals

  • Do not infer electrical pin roles from pin names alone.
  • Do not call a PCB ratline a schematic logical connection.
  • Do not treat explicit no-connects or DNP parts as defects without contradicting evidence.
  • Do not mutate the schematic or assign patterns during a review.

Use document for parser/check output, caller for supplied intent, and heuristic for model prioritization.

Signals

GitHub stars
22
Forks
1
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
schematic-erc-review
Source
github.com/fireostendere/mcp_diptrace