sya-status
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill uses path traversal (
../) to read files in a sibling directory (satisfy-your-agent). This allows the agent to access potentially sensitive data or logic from other skills without explicit isolation.\n- [COMMAND_EXECUTION]: The instructions direct the agent to execute a bundled Python status command. Running local code is a high-privilege action that increases the overall attack surface of the agent environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests content from external sources (sibling directories) without adequate sanitization or boundary markers.\n - Ingestion points: SKILL.md reads files from
../satisfy-your-agent/.\n - Boundary markers: None present to delimit ingested content from skill instructions.\n
- Capability inventory: Local command execution via Python.\n
- Sanitization: No filtering or validation of external content is performed.
Audit Metadata