conflict-detector
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface where instructions could be embedded in the data it processes.
- Ingestion points: Processes proposed task data from artifacts[].path in plan node frontmatter and pending-confirm sibling tasks.
- Boundary markers: No specific delimiters or 'ignore instructions' warnings are defined for the ingested metadata.
- Capability inventory: Executes internal shell scripts (check-l1-files.sh, check-l2-syntactic.sh) using Bash and writes findings to project-local JSONL files.
- Sanitization: The instructions do not specify any validation or sanitization steps for the external metadata before it is used to resolve paths or stored.
Audit Metadata