verify-replay
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external verification reports (
docs/verification/verify-run-report.md). If these reports contain untrusted content, malicious instructions could be injected into the resulting YAML replay manifest. The skill lacks explicit sanitization or boundary markers when interpolating report data into the output file. - [DYNAMIC_EXECUTION]: The skill generates a YAML manifest containing a
runaction and setup/teardown commands (e.g.,bash scripts/dev.sh). This constitutes a script generation surface where the content of the generated script depends on the input file, creating a risk of command injection. - [COMMAND_EXECUTION]: The documentation explicitly instructs agents to execute the commands defined in the generated
replay.ymlfile. This intended behavior allows for arbitrary command execution if the manifest is populated with malicious payload via the input report.
Audit Metadata