gtm-research-outbound
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content such as 10-K filings, earnings call transcripts, and website data. This data is interpolated into prompts to generate sales sequences and briefs, creating a surface for indirect prompt injection if the source material contains malicious instructions.
- Ingestion points: Intelligence sources defined in Step 1a and Step 2, including pasted text or data retrieved via web research (10-K content, transcripts, press releases).
- Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the ingested intelligence data.
- Capability inventory: Access to local file system (
../../profiles/client-profile.md), potential network access for research tools, and subprocess execution for artifact generation. - Sanitization: No evidence of sanitization or escaping of the external content before processing.
- [DYNAMIC_EXECUTION]: The skill specifies the generation of PDF reports using 'Python + reportlab' in the Artifact Generation section. This involves the agent generating and executing Python code at runtime to produce a formatted document.
- [COMMAND_EXECUTION]: The instructions involve searching CRM systems and performing broad web research, which relies on the agent's ability to execute tools and commands to interact with external services and the local file system.
Audit Metadata