confirm
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and designed to prevent unauthorized state-modifying commands by requiring explicit user confirmation.
- [NO_CODE]: There are no executable files, scripts, or remote dependencies included in the skill.
- [INDIRECT_PROMPT_INJECTION]: The skill interprets user requests to generate task summaries, representing a minor ingestion surface. However, the agent is restricted to read-only actions during this phase, mitigating potential risks.
- Ingestion points: SKILL.md (summarizing user task instructions).
- Boundary markers: Missing specific delimiters for user input, though instructions to 'summarize in your own words' act as a natural linguistic boundary.
- Capability inventory: Restricts agent to read-only exploration tools (Read, Grep, Glob) until confirmation.
- Sanitization: None.
Audit Metadata