feature-verify
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill constructs shell commands using string interpolation of variables like {{ BASE_URL }}, {{ ENDPOINT }}, and $BASE_BRANCH. Evidence is found in references/environments.md (health check algorithm) and references/blackbox-testing.md (git diff execution). If these variables are populated with malicious shell metacharacters, it could lead to arbitrary command execution.- [DATA_EXFILTRATION]: The skill handles authentication headers and tokens for API and Log systems. It uses these credentials in curl requests to external hosts defined in environments.md. While the skill includes instructions for redaction and use of fixed test parameters, the handling and transmission of authentication tokens to user-defined endpoints constitute a sensitive data handling surface.- [PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface where external data influences agent behavior.
- Ingestion points: System logs (queried via curl) and git diff output.
- Boundary markers: Missing; log content is analyzed for keywords without strict delimiters.
- Capability inventory: Bash tool (curl, git, gh), WebFetch, and file system access via Read and Grep.
- Sanitization: Absent; the skill does not explicitly escape or validate log signals or file paths before processing.
Audit Metadata