product-marketing
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a vulnerability surface for Indirect Prompt Injection during its automated drafting process.
- Ingestion points: SKILL.md (Step 2) reads repository files including README, landing pages, marketing copy, and package.json.
- Boundary markers: Absent. The instructions do not define delimiters or isolation prompts for external content.
- Capability inventory: SKILL.md (Step 3) performs file system read operations for context gathering and write operations to the .agents/ directory.
- Sanitization: Absent. Information is drafted directly from codebase materials.
- Note: This represents a standard risk for context-aware agents, which is mitigated by the mandatory user review step explicitly defined in the workflow.
- [COMMAND_EXECUTION]: The skill performs routine file system operations, such as checking for file existence, reading project metadata, and writing structured markdown documents. These activities are consistent with the primary purpose and do not involve suspicious execution patterns or privilege escalation.
Audit Metadata