reference-integrity-checker
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and does not utilize any tools, scripts, or system-level commands, which effectively eliminates technical attack vectors related to code execution or privilege escalation.
- [SAFE]: All external links point to the author's official GitHub repository, which is a recognized platform and consistent with legitimate vendor resource patterns for the author 'AIPOCH'.
- [SAFE]: Hardcoded instructions in SKILL.md and references/hard-rules.md explicitly forbid the fabrication of references, DOIs, and PMIDs, addressing the risk of hallucinated or malicious academic data.
- [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it processes untrusted user-provided manuscript drafts and rebuttal text (Ingestion points: SKILL.md). There are no boundary markers or delimiters specified to isolate these inputs. However, the capability inventory is empty, as the skill lacks allowed-tools, network access, or file-system write permissions, and no sanitization is present. Consequently, the impact of any potential injection is negligible.
Audit Metadata