Audit History

Most static detections are Markdown formatting, expected Claude paths, test syntax, or documentation links. Multiple executable templates directly interpolate positional arguments into shell text, creating command-injection risk. Predictable temporary files and repeated Bash(*) grants add avoidable exposure, but no prompt injection or covert exfiltration intent was found. Static review was capped at 400/429 representative findings; omitted static matches are unconfirmed, so automatic publishing stays disabled until manual review.

The static findings are Markdown examples for Claude Code slash command syntax, not runnable install or auto-execution logic. No prompt injection, data exfiltration, or malicious intent was found in the reviewed files.

The static findings are predominantly false positives from Markdown documentation and slash command examples, not executable skill code. No prompt injection attempt, credential exfiltration, or install-time execution behavior was found. The main residual risk is instructional: some examples show broad Bash permissions and direct shell interpolation of user arguments.

This official Anthropic skill contains documentation and examples for building Claude Code slash commands. Static indicators reference bash context, file references, and network examples in documentation. No malicious behavior or prompt injection attempts were confirmed.