cawplan-testplan-layout
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data which could potentially influence command execution.
- Ingestion points: Untrusted data enters the agent context via the
argument-hintwhich accepts product portal links,product_id,version_id, andticket_id(SKILL.md). - Boundary markers: There are no explicit delimiters or instructions provided to the agent to escape or ignore malicious payloads embedded within the user-provided IDs or links.
- Capability inventory: The skill is configured with
allowed-tools: Bashand uses it to execute multiple CLI commands, includingcawplan qa-insights testrail plan previewandplan execute, which interpolate the user-provided variables directly into the command string (execution.md §Preview). - Sanitization: The instructions do not define any validation or sanitization logic to verify that the provided IDs are numeric or conform to expected formats before being passed to the shell.
Audit Metadata