literature-review
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill contains no malicious code, obfuscation, or evidence of data exfiltration. All instructions are focused on legitimate academic research tasks.
- [NO_CODE]: The skill is comprised solely of markdown instructions and YAML frontmatter. It does not include scripts, binaries, or automated configuration changes that could execute commands on the host system.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external academic research papers, which creates a potential surface for indirect prompt injection.
- Ingestion points: External data is ingested via "papers supplied" or through a defined "search scope" (SKILL.md).
- Boundary markers: The instructions do not define specific delimiters for separating the agent's instructions from the content of the research papers.
- Capability inventory: The skill's primary capabilities are text extraction, thematic organization, and synthesis.
- Sanitization: The structured extraction process (Step 3), which requires the agent to identify specific methodological details and claims, serves as a functional filter that reduces the likelihood of the agent obeying adversarial instructions embedded in the source text.
Audit Metadata