attention-frontier
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a coordination tool that aggregates project status from thread history. It uses standard platform-provided tools such as list_threads and read_thread to gather information.
- [SAFE]: Security is enhanced by frontmatter configuration that disables autonomous model invocation (disable-model-invocation: true) and implicit execution (allow_implicit_invocation: false), ensuring the agent only acts when directed and within strict constraints.
- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it reads external conversation content. However, the risk is minimal because the skill's primary function is observation and prioritization, and it lacks the capabilities (like network access or command execution) required to exploit such an injection.
- [EXTERNAL_DOWNLOADS]: The skill references official OpenAI documentation for deep links (learn.chatgpt.com), which is a trusted, well-known service.
Audit Metadata