pippit-skill
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements safe security practices for credential management, restricting cache file permissions via
os.chmodto owner-only read/write (0o600). Network operations are limited to the platform's official domains, representing standard vendor operations without risk of unauthorized data exfiltration or third-party remote code execution. - [INDIRECT_PROMPT_INJECTION]: The skill processes external responses from an AI generation endpoint, presenting a standard indirect injection surface.
- Ingestion points: Untrusted text sequences and questionnaires are fetched from the remote thread via
scripts/nest/get_thread.py. - Boundary markers: Data is passed as structured JSON without individual prompt-level containment delimiters.
- Capability inventory: The file operations are confined to downloading asset files in
scripts/nest/download_results.pywith no execution paths for shell commands or dynamic scripts. - Sanitization: Responses are validated via structured JSON decoding without custom character filters.
Audit Metadata