content-publish-check
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user-provided content (blog posts, social posts, drafts) as its primary input while possessing the capability to publish content.\n
- Ingestion points: The skill reads user drafts for proofreading, editing, and fact-checking in Step 1 and Step 3.\n
- Boundary markers: There are no instructions to use delimiters or ignore embedded instructions when reading the draft content, which could allow a malicious draft to influence the agent's behavior during the review process.\n
- Capability inventory: The agent has the capability to write and publish content via the
write_posttool and other related skills mentioned in the Related Skills section.\n - Sanitization: The instructions do not specify any sanitization or validation of the input draft to prevent malicious commands embedded within the text from being executed by the agent.\n
- Mitigation: The risk is mitigated by Step 6, which requires explicit human owner approval before any content is published.
Audit Metadata