prestidigitation
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONINGNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user data for tasks such as reformatting, sorting, and encoding/decoding, which creates a potential surface for indirect prompt injection.
- Ingestion points: User-supplied artifacts including text strings, tables, and blobs described in the 'When To Use' section.
- Boundary markers: The procedure requires the agent to restate 'no-touch boundaries' before execution, providing a manual check but no technical isolation.
- Capability inventory: The skill relies on the agent's default toolset as no custom scripts are provided.
- Sanitization: There are no explicit instructions for sanitizing or escaping the content being transformed.
- [METADATA_POISONING]: The skill metadata contains tags such as 'actuation', 'access', and 'execution' that directly contradict the primary description and guardrails which explicitly state the skill does not perform actuation or touch infrastructure.
- [NO_CODE]: The skill consists entirely of markdown instructions and does not include any executable code, scripts, or binary files.
Audit Metadata