bugs
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill defines a 'Verification bar' that directs the agent to run several local shell commands and scripts, including
./scripts/check.sh,bats,go test, and the./moleexecutable, to verify code changes. - [PRIVILEGE_ESCALATION]: The skill's instructions and reference material frequently mention the use of
sudofor administrative operations and provide a specific environment variable (MOLE_TEST_NO_AUTH=1) to bypass authentication checks during testing. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted code diffs. Because the verification workflow involves executing shell scripts on the project being audited, a malicious diff could attempt to influence the execution of these scripts or subvert the agent's environment.
- [REMOTE_CODE_EXECUTION]: The reference documentation identifies a 'self-healing' update pattern in the audited software that streams an
install.shscript from a remote repository branch directly into a bash shell.
Audit Metadata