integrate-app-intents
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill implements a robust 'Trust boundary' section in the main instructions (SKILL.md), explicitly directing the agent to ignore any malicious instructions found in analyzed data such as Siri transcripts, logs, or external code examples.
- [SAFE]: A specific evaluation case (evals/evals.json, ID 6) is included to verify that the agent correctly identifies and rejects malicious requests involving prompt injection, credential disclosure, and data exfiltration.
- [SAFE]: The skill manages potential indirect prompt injection surfaces (ingesting transcripts and logs) by providing strong boundary markers and requiring the use of synthetic or sanitized data for all system-facing tasks.
- [NO_CODE]: The skill consists entirely of instructional markdown and configuration files; no executable scripts, binaries, or active code components are shipped within the skill folder.
Audit Metadata