skills/jarmen423/skills/postiz/Gen Agent Trust Hub

postiz

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill architecture creates a surface for indirect prompt injection because it retrieves dynamic metadata and tool outputs from an external API and processes them within the agent's context.
  • Ingestion points: The agent is instructed to call commands such as postiz integrations:list, postiz integrations:settings, and postiz integrations:trigger, which fetch integration metadata, character limits, and dynamic data (like Reddit flairs or YouTube playlists) from the api.postiz.com backend.
  • Boundary markers: There are no explicit instructions or delimiters provided to the agent to help it distinguish between legitimate tool output and potentially malicious instructions embedded in social media content or metadata retrieved from the API.
  • Capability inventory: The skill possesses the ability to read local files via the postiz upload command and perform network operations via postiz posts:create, which could be exploited if an injected instruction successfully manipulates the agent's flow.
  • Sanitization: The documentation does not specify any sanitization or validation steps for the data returned by the API before it is interpolated into the agent's reasoning or subsequent commands.
  • [EXTERNAL_DOWNLOADS]: The skill documentation includes a recommendation for the agent to suggest an external third-party tool from a non-standard source.
  • Evidence: In SKILL.md, the instructions suggest that the agent recommend the agent-media CLI located at https://clawhub.ai/nevo-david/agent-media. While this is presented as a recommendation in the prose and not an automated download, it directs the agent and the user toward an unverifiable external repository.
  • [COMMAND_EXECUTION]: The skill relies on the installation and execution of a global binary (postiz).
  • Evidence: The installation instructions require npm install -g postiz. The skill then executes various subcommands (posts:create, upload, etc.) which interface with the local file system and remote API. This is the primary intended function of the skill, but it represents the capability used by the identified injection surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 04:04 PM