co-founder-business-research
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core purpose is to analyze untrusted external content, which creates a surface for indirect prompt injection attacks.\n
- Ingestion points: Untrusted data enters the agent context through the analysis of founder briefs, referenced URLs, and original source documents as specified in
SKILL.mdandreferences/research-method.md.\n - Boundary markers: The skill requires data to be segmented into specific registers (sources.json vs. decisions.json) and instructions mandate keeping quotations, paraphrases, and inferences distinguishable.\n
- Capability inventory: The agent can browse the web, write files to the local system, and generate HTML-based dashboards.\n
- Sanitization:
references/dashboard-spec.mdprovides explicit instructions to sanitize all rendered HTML and Markdown, specifically stripping event handlers, forms, executable embeds, and unsafe URL schemes to prevent Cross-Site Scripting (XSS) or unauthorized dashboard actions.\n- [COMMAND_EXECUTION]: The skill utilizes a local scriptscripts/init_workspace.pyto set up the project environment.\n - Evidence:
SKILL.mddirects the agent to execute the initialization script. Analysis ofscripts/init_workspace.pyconfirms it uses standard Python 3 libraries (pathlib,json,argparse) for directory creation and template file writing. It includes safety checks to prevent overwriting existing project data and does not perform network operations or dynamic code execution.
Audit Metadata