Review Evidence

Compare a register map with its source and show only the missing or contradictory details that need your answer.

GROUP 02Review register mapsGitHub source ↗

WHAT YOU GET

A clear result you can inspect

It writes `report.json`. This short report shows only the details that the manual does not settle. For example, one unclear byte-order note that affects 20 points becomes one question. Values stated clearly in the manual do not appear.

HOW IT RUNS

01 · GUIDESKILL.mdprocess + safety

02 · EXECUTEscripts/run.pybundled local Python

Compares the map with the source manual. It checks every value against the original file or page reference. Clear matches need no action. Missing byte order and missing packed-bit numbering remain separate questions. It groups each shared problem by source section, so you do not review the map one row at a time.

SYNTHETIC report.json

Review one shared question, not 20 rows

The report keeps clear points out of the review and groups points that need the same answer.

20points with an unclear byte order

1question for your approval

review_status: blockedblocking_decisions: 1

WHERE IT FITS

Used in these workflows

Extract PDF register evidenceExtract traceable rows and grouped exceptions from a PDF without performing a complete source-map review.view workflow →

EXAMPLE REQUEST

Ask for the result in your own words.

Use this example as a starting point. Replace the file names and details with your own.

CODEX$review-evidence

CLAUDE CODE/modbus-skills:review-evidence

CURSOR + OTHER CLIENTSUse the client’s native skill invocation.

$review-evidence

Please compare `normalized-map.json` with its original file and page references. Show me only the details that are missing or do not match, and put related questions together.

GitHub source ↗