cleanup-deleted
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local TypeScript script (
cleanup_deleted.ts) located in the plugin root directory using thebunruntime environment within a Bash block.- [INDIRECT_PROMPT_INJECTION]: The skill contains 'Self-Evolving' instructions and a 'Post-Execution Reflection' section that directs the agent to modify theSKILL.mdfile if execution errors occur or if the underlying tool interface changes. This creates a potential surface where external tool output could influence the agent's future instructions through automated file updates. - Ingestion points: Command output from
bun "$SCRIPT"(SKILL.md) - Boundary markers: None present for the output processing
- Capability inventory: Bash execution, Read/Write access to the skill file itself
- Sanitization: Not applicable as the agent is instructed to manually update the instructions based on observation.
Audit Metadata