postiz
Warn
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script
scripts/sync-facts.mjsuses theeval()function to parse a data structure from a source file in a separate repository. This is used to synchronize pricing and limits data between the application code and the documentation. - [CREDENTIALS_UNSAFE]: The utility script
scripts/screenshots/capture.mjscontains hardcoded credentials (EMAIL = 'demo@acme.test'andPASSWORD = 'DocsDemo!2345'). While these are intended for a local demo instance used to generate documentation screenshots, hardcoding credentials in repository scripts is an unsafe practice. - [INDIRECT_PROMPT_INJECTION]: The skill provides automated features that ingest untrusted data from external sources, creating a significant vulnerability surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the agent context through RSS feeds (
general/automations/auto-post-rss.mdx) and YouTube video transcripts or captions processed for the clipping feature (general/clipping.mdx). - Boundary markers: Explicit warnings are present in the documentation (
public-api/clipping/get.mdx) and MCP tool schemas (mcp/tools.mdx), instructing the agent to treat ingested text (such as clip titles and post text) as content for the user and never as instructions. - Capability inventory: The skill possesses the capability to schedule and publish posts (
schedulePostToolinmcp/tools.mdx) and perform network operations via the Postiz API. - Sanitization: No evidence of automated sanitization, filtering, or escaping of external content is described before the data is interpolated into the agent's context or drafted for social media posts.
Audit Metadata