sales-meeting-recap
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data from call notes and transcripts.
- Ingestion points: User-provided sales notes, transcripts, or descriptions in SKILL.md.
- Boundary markers: Lacks explicit delimiters or markers to separate untrusted input from system instructions.
- Capability inventory: Capabilities are limited to generating email text and mutual action plans; no network, file system, or subprocess tools are used.
- Sanitization: Includes an 'Anti-fabrication rule' to prevent invented commitments, but no specific sanitization against embedded prompt instructions.
- [NO_CODE]: The skill contains no executable scripts or binaries. All logic is implemented via natural language instructions and reference documentation.
Audit Metadata