think-random-frameworks
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted user data ('topic') which is interpolated into the agent context to drive framework application.\n
- Ingestion points: User topic input as defined in
SKILL.mdand processed via the logic inreferences/engine.md.\n - Boundary markers: Absent. The instructions do not specify the use of delimiters (e.g., XML tags) or 'ignore embedded instructions' warnings for the user-provided topic.\n
- Capability inventory: The skill reads various local markdown files (instructions, templates, and corpus data) from its repository paths.\n
- Sanitization: No input validation, escaping, or filtering is implemented for the external content before processing.\n- [SAFE]: The skill uses local file access and external URLs that are consistent with the author's verified infrastructure and intended functionality (product-on-purpose GitHub organization).\n- [SAFE]: No evidence of obfuscation, hardcoded credentials, persistence mechanisms, or unauthorized command execution was found.
Audit Metadata