spec
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (MEDIUM): The
/spec --import <url>and/spec --synccommands allow the agent to fetch content from any user-provided URL. This data is downloaded into the local project structure without source validation. - COMMAND_EXECUTION (MEDIUM): The skill utilizes the
Bashtool to handle file operations and network fetching (cloning/syncing). If the URL or the resulting directory structure is maliciously crafted, it could lead to command injection or unauthorized file system operations. - PROMPT_INJECTION (LOW): The skill is vulnerable to indirect prompt injection through external specifications. Ingestion points: Remote specification files are fetched and saved to
docs/specs/. Boundary markers: There are no explicit instructions to the agent to ignore or treat the imported content as untrusted data when reading them to 'suggest initial TASKs'. Capability inventory: The agent has access toBash,WebFetch,Write, andEdittools. Sanitization: No sanitization or safety checks are performed on the imported markdown content before it is parsed and acted upon by the agent.
Audit Metadata