chemical-structure-converter
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill package includes a metadata file named
chemical-structure-converter_audit_result_v4.jsoncontaining self-referential claims about the skill's safety status, such as 'security: PASS'. This is a pattern used to influence automated security analyzers. - [INDIRECT_PROMPT_INJECTION]: The skill documents the capability to process batch inputs via CSV or plain text lists, which establishes an ingestion point for untrusted data. This creates a surface for indirect prompt injection if the agent interprets instructions hidden within the chemical datasets.
- Ingestion points: User-provided chemical identifiers and batch input files (CSV/text) mentioned in
SKILL.md. - Boundary markers: The instructions lack delimiters or warnings to ignore instructions embedded in the chemical data.
- Capability inventory: No dangerous functions like network exfiltration or file writing are currently implemented in the skill's script.
- Sanitization: No validation or sanitization of input data for non-chemical content is present.
Audit Metadata