positioning-messaging
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources including user-provided transcripts and attached images. While this represents a potential injection surface, it is central to the skill's intended purpose and mitigated by quality gates.\n
- Ingestion points: The skill parses attached images in Step 5 and ingests customer evidence such as win/loss transcripts and quotes in Phase 1 (SKILL.md).\n
- Boundary markers: The skill utilizes a 6-phase discovery process and a 7-point verification gate to ensure output consistency.\n
- Capability inventory: The skill writes session metadata to the local file
/context/skill-sessions.md(SKILL.md).\n - Sanitization: A mandatory quality gate in Phase 6 checks for jargon and resolves "vision flags" identified during image parsing.\n- [SAFE]: The skill maintains session persistence by appending data to a local log file.\n
- Evidence: Phase 7 (Learning Close) appends a YAML entry to
/context/skill-sessions.md. This is a benign logging mechanism limited to the local workspace.
Audit Metadata