← ClaudeAtlas

akcli-schematic-reviewlisted

Run a structured, severity-ranked design review (審查) of an Altium .SchDoc or KiCad .kicad_sch schematic using the `akcli` CLI. Use this skill whenever the task involves: reviewing a schematic before release, prototype, or tape-out; auditing ERC, power-rail, or BOM check results; spot-checking a netlist against datasheet pin functions; cross-checking an MCU pinmap against an expected pin table (CSV/JSON/DTS-derived); reviewing what changed between two schematic revisions and whether anything regressed; running the `akcli review` engine (signal/validation/PCB/EMC/gerber detectors) and interpreting its confidence-graded findings, trust summary, and EMC risk score; or writing a hardware review report with Blocker/Major/Minor/Info/Question findings. Triggers on keywords: design review, schematic review, circuit review, 審查, hardware review, ERC audit, power tree, decoupling, pull-up, strap pin, boot pin, floating input, connector pinout, netlist check, pinmap cross-check, revision diff, regression, review report, s
tipoLi5890/akcli · ★ 4 · Data & Documents · score 74
Install: claude install-skill tipoLi5890/akcli
# akcli-schematic-review — a severity-ranked schematic review protocol on top of `akcli` This skill is the **review protocol**: what to run, in what order, how to interpret the caveats, which electrical checks the tool cannot do for you, and how to write the report. For basic `akcli` mechanics (input formats, read/analyze/draw usage, config discovery, exit codes, error format) **see the akcli-circuit-design skill** — do not re-derive them here. A review is **strictly read-only**. Never modify the schematic under review, and never "fix" it to match an expected table — report the divergence instead. ## Review doctrine - **Evidence or it didn't happen.** Every finding — and every "checked, OK" — must cite the exact `akcli` command you ran and paste the relevant output line(s). Never assert a defect or a pass from memory or from reading the file listing alone. - **"0 findings" is not "clean".** Read the metadata header `check` prints (passive-pin ratio, No-ERC suppressed count, unnamed-net count, frac coords) before concluding anything. - **The schematic is authoritative; expected tables and datasheet notes are advisory.** - **Confidence is part of the finding.** `review` findings carry an explicit confidence: `deterministic` (recomputable geometry/math — trust it), `datasheet_backed` (carries the PDF sha256+page — trust it, cite the page), `heuristic` (name/prefix inference — YOU adjudicate before reporting above Minor), `llm_reviewed` (an accepted observation