paper-title-abstract-keywords
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting untrusted research data that is subsequently utilized in prompt generation and file-writing tasks.
- Ingestion points: The skill collects "Research theme and questions", "Core findings", and "Theoretical positioning" (found in the 'Startup' section of SKILL.md) which are supplied by the user or retrieved from the 'ta-research-AFP' context.
- Capability inventory: The skill uses a 'Write' tool to save generated content to a file named 'title_abstract_keywords.md' in the current project directory.
- Boundary markers: The instructions do not define specific delimiters or instructions to the agent to disregard potential commands embedded within the provided research summaries or findings.
- Sanitization: There is no evidence of input filtering or validation to prevent the inclusion of malicious instructions or script content within the user-provided data being written to the filesystem.
Audit Metadata