citation-record
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script (
python3 scripts/score.py --file draft.json) to validate the citation records. This is a standard operational step within the context of the skill's purpose. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing untrusted data from 'buyer questions' which could potentially contain malicious instructions.
- Ingestion points: The skill processes external 'buyer questions' as the primary input (SKILL.md).
- Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the buyer question content as untrusted data or to ignore embedded instructions.
- Capability inventory: The skill has the capability to execute Python scripts via the shell (SKILL.md).
- Sanitization: There is no mention of sanitization, filtering, or validation of the buyer question content before it is used to generate the citation record or processed by the scoring script.
Audit Metadata