egm-resolve
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill processes deficiency reports from external sources and incorporates that data into file modifications and chat responses, which could allow instructions in the review data to influence agent behavior.
- Ingestion points: The FindingsList input parameter is processed in Step 1 of SKILL.md.
- Boundary markers: The instructions use markdown templates but do not include explicit delimiters or 'ignore' instructions for the content within the findings.
- Capability inventory: The skill utilizes file-writing and file-editing tools for generating reports and applying fixes in SKILL.md (Steps 3 and 4).
- Sanitization: There is no evidence of content sanitization or validation for the ingested findings data.- File System Interaction: The skill performs read and write operations on the local file system to process documents and generate reports.
- It reads the user-specified TargetFile and creates a goldfish_review_report.md artifact.
- These actions are intended for the skill's primary purpose and are mitigated by an interactive resolution gate requiring user confirmation before any modifications are finalized.
Audit Metadata