context-building
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data which could contain malicious instructions.
- Ingestion points: External data enters the agent context via 'Mode 3: Call Recording Capture' (call transcripts/meeting notes) and 'Mode 4: Feedback Loop' (CSV or email sequencer exports like Instantly) in
SKILL.md. - Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to isolate the untrusted external content from the agent's logic.
- Capability inventory: The skill possesses file-writing capabilities, specifically creating and appending to files within the
claude-code-gtm/context/directory as described inSKILL.mdandreferences/context-schema.md. - Sanitization: No explicit sanitization, validation, or escaping rules are defined for the extracted data.
- Mitigation: The skill includes a mandatory verification step in Mode 3, Step 3 ('Present extracted signals to the user for confirmation'), which acts as a manual guardrail before data is committed to the context file.
Audit Metadata