enterprise-carbon-credit-rating
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from customer materials (Word, Excel, PDF, OCR text) to extract carbon credit rating fields. This workflow creates an indirect prompt injection surface where adversarial content in the uploaded documents could attempt to influence the agent's extraction behavior or scoring logic. The skill mitigates this by implementing a mandatory confirmation gate where the user must review and approve all extracted identity fields and industry parameters before any scoring or report generation occurs.\n
- Ingestion points: Customer-supplied documents including Word, Excel, PDF, and OCR-extracted text files mentioned in
SKILL.md.\n - Boundary markers: Not explicitly present in the data extraction phase, but a human review gate is required before scoring.\n
- Capability inventory: Execution of local Python scripts (
scripts/score_company.py,scripts/build_report.py) and file system write operations (rating_result.json,企业碳资信评级报告.docx).\n - Sanitization: The user is required to confirm all extracted values in a summary table before final processing.\n- [COMMAND_EXECUTION]: The skill uses local Python scripts to perform deterministic scoring and build Word documents. These commands are documented in
SKILL.mdand are necessary for the skill's primary functionality.
Audit Metadata