Trim
Warn
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the bun runtime to run a local utility script (LIFEOS/TOOLS/ProposalGC.ts). It also performs various git and wc commands for version control and file measurement.\n- [DATA_EXFILTRATION]: The skill accesses, reads, and writes to sensitive files including OPERATIONAL_RULES, PRINCIPAL_MEMORY.md, and DA_MEMORY.md. It instructs the agent to commit these changes to a private Git repository at ~/.config/LIFEOS/USER.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the data it processes.\n
- Ingestion points: Reads instructions and data from OPERATIONAL_RULES, CLAUDE.md, and memory files.\n
- Boundary markers: Employs a 'Safety gate' to verify that no directives are lost, but lacks markers to isolate instructions within the processed data.\n
- Capability inventory: Possesses the ability to write to system files, execute shell commands, and run scripts via bun.\n
- Sanitization: Does not perform sanitization of the file content before processing or writing.
Audit Metadata