create-faceless-channel
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external APIs, RSS feeds, and web scrapers to generate video scripts and metadata. This introduces a vulnerability surface where malicious content within those sources could attempt to influence the agent's behavior during content generation.
- Ingestion points: Untrusted data enters the context during the 'Discover' and 'Capture' stages defined in
SKILL.mdandreferences/pipeline-contract.md, which involve fetching content from documented APIs, feeds, or browser scrapers. - Boundary markers: The skill includes mitigation instructions in
references/pipeline-contract.md, such as requiring the agent to "Bind claims to source locations" and "Keep a separate statement for inference or opinion" to prevent verbatim copying of potentially malicious source instructions. - Capability inventory: The skill possesses the capability to write files via
scripts/scaffold_channel.pyand perform network operations to publishing schedulers (Postiz) and alert services (AgentMail). - Sanitization: Guidelines in
references/pipeline-contract.mdmandate rejecting scripts with malformed punctuation, source navigation instructions copied as narration, or generic praise, which serves as a basic sanitization layer for external content.
Audit Metadata