skill-sync
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes metadata from other files in the
skills/directory and propagates it into documentation. While this is the intended purpose, it establishes a data flow where instructions from one skill's metadata could reach the agent's primary instruction files.\n - Ingestion points: Reads
metadata.descriptionandmetadata.auto_invokefrom allSKILL.mdfiles in the repository.\n - Boundary markers: Uses HTML comments to mark the start and end of generated sections in
AGENTS.md.\n - Capability inventory: Requires
Read,Edit,Write, andBashpermissions to update files.\n - Sanitization: Metadata content is extracted and placed into markdown tables without additional escaping or instruction-boundary wrappers.\n- [COMMAND_EXECUTION]: Employs bash scripts (
sync.sh,sync_test.sh) for file system operations and string manipulation viaawk,sed, andfind. All operations are performed locally within the repository scope.
Audit Metadata