moai-alfred-todowrite-pattern
Pass
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill implementation patterns interpolate untrusted data directly into agent prompts. For example, in
reference.md, a user request is embedded into a prompt for a 'Plan' agent:prompt=f"Create structured plan for: {user_request}".\n - Ingestion points: The
user_requestvariable and task-related fields likecontent.\n - Boundary markers: No boundary markers or specific delimiters are used to isolate the untrusted input from the rest of the prompt.\n
- Capability inventory: The skill possesses
BashandTodoWritecapabilities, enabling potential file-system modifications or command execution.\n - Sanitization: No evidence of sanitization, filtering, or escaping of input data is present in the provided patterns.\n- [COMMAND_EXECUTION]: The skill's metadata in
SKILL.mdexplicitly enables theBashtool. Furthermore, the markdown body includes an example of using Bash to modify local files (e.g., appending content to the skill's own file usingcat >>), which could be abused if an injection occurs.
Audit Metadata