paw-pa-generation
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/export_proposal.pyscript executes thepandoccommand-line utility to convert proposal drafts into various export formats like PDF and DOCX. This execution is performed securely using thesubprocess.runmethod with arguments passed as a list, which prevents command injection vulnerabilities. - [PROMPT_INJECTION]: The skill ingests data from external artifacts, specifically
brief.mdandresearch-dossier.html, to populate the proposal content. This introduces a surface for indirect prompt injection where untrusted content from these files could attempt to influence the agent's behavior. - Ingestion points: The skill reads artifacts from
{run-folder}/brief.mdand{run-folder}/research-dossier.htmlduring the assembly phase. - Boundary markers: The assembly instructions do not specify explicit delimiters or "ignore instructions" markers for the artifact content.
- Capability inventory: The skill has the capability to read and write files within the project workspace and execute a local helper script for document rendering.
- Sanitization: The accompanying export script provides HTML escaping for content before it is rendered into the final branded HTML output.
Audit Metadata