strategic-gifting
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection by ingesting untrusted data from external sources.
- Ingestion points: Recipient research data (hobbies, LinkedIn posts, podcast appearances) defined in
templates/gift-brief.mdandreferences/giftology-ruhlin.md. - Boundary markers: None identified; instructions do not specify delimiters for external content.
- Capability inventory: Uses
scripts/check-output.pyfor local text validation; the agent likely possesses file-writing and tool-use capabilities to finalize the gifting brief. - Sanitization: No sanitization, escaping, or filtering of the external research content is described before it is interpolated into the gift note drafts.
- [COMMAND_EXECUTION]: The skill requires the execution of a local Python script
scripts/check-output.pyto validate deliverables. While the script performs benign regex-based content checks, it represents a point where the agent is instructed to run code locally.
Audit Metadata