social-creative-builder
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from users (ideas, source articles, scraped content). It identifies this as a potential attack surface and mitigates risk by instructing the agent that such content cannot be used to approve product claims, waive disclosures, or modify platform settings. Hand-off summaries and claims handling are restricted to an internal ledger.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads project-specific configuration files (narrative, claims, and channel projections) stored in the local memory directory. It uses internal helper scripts (
tavily.py,bluesky.py) for fact-checking and norm-verification, which are aligned with the skill's primary purpose. No evidence of harvesting sensitive system credentials or exfiltrating data to unauthorized endpoints was found. - [COMMAND_EXECUTION]: The skill interacts with local scripts (
registry-events.py) to submit authorized proposals. These interactions are structured and do not involve the execution of arbitrary shell commands or user-supplied code injection. - [SAFE]: The skill follows established best practices for agent-skill architecture, including the use of specific read/write scopes and human-in-the-loop requirements for content publishing.
Audit Metadata