persistence-audit
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script executes a wide array of macOS system utilities including
launchctl,PlistBuddy,sfltool,systemextensionsctl,codesign,security,kmutil,lsof, andsocketfilterfw. These are used purely for diagnostic enumeration. - [PRIVILEGE_ESCALATION]: The documentation in
SKILL.mdexplicitly instructs the user/agent to run the script withsudoto access root-only sections of the persistence audit, such as system-wide LaunchDaemons and thesecurity authorizationdb. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (system configuration files, service labels, and command outputs). Since this data could theoretically be manipulated by malicious software on the host machine, it represents an indirect prompt injection surface where the agent might interpret system metadata as instructions.
- Ingestion points:
persistence-inventory.shreads plists from/Library/LaunchDaemons,/Library/LaunchAgents, and various command outputs likesfltool dumpbtm. - Boundary markers: Output is grouped by sections with text headers, but no explicit instruction-ignoring delimiters are used to separate system data from agent instructions.
- Capability inventory: The skill uses
BashandReadpermissions to execute system tools and read system files. - Sanitization: No sanitization or escaping is performed on the gathered metadata before printing it to the console.
Audit Metadata