The Agent Skills Directory

[SAFE]: The skill implements a 'read-only' diagnostic approach, specifically instructing the agent not to modify state or apply fixes even if requested by the user, which prevents state-modifying prompt injection attacks.
[SAFE]: Includes a dedicated security step (Phase 3, Step 1: Scrub sensitive content) that mandates the identification and redaction of API keys, tokens, and passwords from the final report output.
[SAFE]: The tool configuration in the frontmatter is limited to Read, Grep, and Glob, effectively preventing unauthorized network access, command execution, or file system modifications.
[SAFE]: No obfuscation techniques, malicious prompt injection patterns, or remote code execution vectors were detected.

forensics