proposal
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from multiple external and potentially untrusted sources to generate business proposals. \n
- Ingestion points: Data is gathered via tools like
list_findings,enrich_company,enrich_person, andsearch_knowledge_base(documented inreferences/octave-tool-reference.md), which pull from CRM records and call transcripts. \n - Boundary markers: The instructions lack specific boundary markers or 'ignore instructions' directives to prevent the agent from following malicious commands that might be embedded in ingested deal notes or transcript data. \n
- Capability inventory: The agent has the ability to write HTML files to the local file system (in
.octave-proposals/) and is directed to invoke a bash script (export-pdf.sh) for PDF conversion. \n - Sanitization: No explicit sanitization or validation of the ingested content is described in the provided references before it is interpolated into proposal documents. \n- [COMMAND_EXECUTION]: The skill provides instructions for executing a local bash script (
export-pdf.sh) to handle PDF generation (found inreferences/delivery-summary.md). While the script is a local utility, executing shell scripts based on agent-generated files involves risk if those files contain unsanitized data from external sources.
Audit Metadata