scheduling-and-queue
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or vulnerabilities were detected. The skill follows best practices for AI agent safety, including:
- [PROMPT_INJECTION]: Mitigation of indirect prompt injection is integrated into the core workflow. The skill instructs the agent to treat all post content, documents, and fetched data as passive data rather than actionable commands. It specifically warns against expanding scope based on text found within processed content.
- [DATA_EXFILTRATION]: Data handling is restricted to the authorized WoopSocial MCP. The skill does not access local sensitive directories (~/.ssh, ~/.aws) or environment variables.
- [COMMAND_EXECUTION]: Side-effectful operations (scheduling, publishing, deleting) are gated by a mandatory 'confirmation preview' step. This human-in-the-loop requirement ensures that the agent cannot be coerced into performing unauthorized actions without user awareness.
- [PROMPT_INJECTION]: Ingestion points: Skill processes user-supplied text and media, as well as content from documents or external tools (SKILL.md, Step 1). Boundary markers: Implements a mandatory visual preview of all planned actions before execution (references/safety-and-confirmation.md). Capability inventory: High-impact capabilities including post creation and deletion via WoopSocial MCP. Sanitization: Explicit instructions to ignore instructions found within data content and rely solely on direct chat instructions.
Audit Metadata