Audit History

All 400 representative static findings are false positives caused by RBS reference signatures, documentation syntax, and bundled Zeitwerk example source. The skill instructions generate and validate type signatures; no evidence shows secret collection, network transfer, shell execution, identity changes, or prompt-injection intent. Static review was capped at 400/1166 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

The reported static findings are false positives from bundled RBS reference signatures, Ruby standard library API declarations, Markdown formatting, and vendored Zeitwerk examples. No evidence of prompt injection, credential theft, malicious network use, or hidden command execution was found in the reviewed context. Static review was capped at 400/1166 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

AI review found the static alerts are scanner noise from Markdown code formatting, RBS signatures, and bundled reference Ruby source. No prompt-injection instructions, credential exfiltration intent, or active install/runtime behavior were found. Command guidance is limited to fixed RBS validation tools.

AI review found the static alerts are scanner noise from Markdown code formatting, RBS signatures, and bundled reference Ruby source. No prompt-injection instructions, credential exfiltration intent, or active install/runtime behavior were found. Command guidance is limited to fixed RBS validation tools.

Static analysis reported many severe patterns, but representative review shows they are mostly RBS reference signatures, Markdown grammar examples, and bundled Ruby examples used as documentation. No malicious intent or prompt injection was found. The remaining real risks are intended filesystem edits and local validation commands, so publication is acceptable with a medium-risk warning.

Static findings (1979) are false positives. The scanner misinterprets reference documentation files as executable code. All external_commands findings are markdown code blocks, blocker findings are RBS type signatures, and network findings are documentation links. The skill only reads Ruby files and generates RBS signatures - no network calls, credential access, or dangerous capabilities exist.