skill-factory
Pass
Audited by Gen Agent Trust Hub on Feb 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill architecture creates a surface for indirect prompt injection (Category 8) by processing external research data to generate new skill instructions.\n
- Ingestion points: External research materials are processed by
scripts/format_skill_research.pyand then synthesized intoSKILL.mdduring the 'write' phase of the workflow.\n - Boundary markers: Absent. While the workflow follows a logical sequence, there are no specific prompt delimiters mentioned to prevent the LLM from obeying instructions embedded in the research text.\n
- Capability inventory: The orchestrator utilizes the
SlashCommandtool and can write files to the file system to initialize and modify skill structures (documented inSKILL.md).\n - Sanitization: Present. The
scripts/format_skill_research.pyscript uses regular expressions to remove UI-specific artifacts from research materials, providing structural but not behavioral sanitization.
Audit Metadata