docforge-dashboard
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data from the repository, which creates a surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the agent context through
.docforge/manifest.json,.docforge/provenance/*.jsonsidecars, document frontmatter, and Markdown bodies indocs/**(SKILL.md). - Boundary markers: The instructions explicitly state that ingested text is treated as inert repository data and never followed as instructions or configuration (SKILL.md).
- Capability inventory: The skill has the ability to execute
npm,node,python3,git, and the system's default browser. It can also perform file writes within the.docforge/directory (SKILL.md). - Sanitization: The skill performs structural validation of JSON metadata and provenance records against supported schema versions and ensures all file paths resolve within the repository boundaries (SKILL.md).
- [COMMAND_EXECUTION]: The skill uses
python3,node, andnpmto execute its internal runtime scripts (dashboard.py and dashboard.js) and manage the dashboard's local environment.
Audit Metadata