pp-prediction-goat

Pass

Audited by Gen Agent Trust Hub on Jul 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the prediction-goat-pp-cli binary using npx from the @mvanhorn/printing-press NPM package and go install from the github.com/mvanhorn/printing-press-library repository. These resources are hosted on the vendor's infrastructure.
  • [COMMAND_EXECUTION]: The skill's primary functionality is achieved by executing the prediction-goat-pp-cli binary through the shell, using the Bash tool to perform data retrieval and management tasks.
  • [DATA_EXFILTRATION]: The CLI tool supports a --deliver webhook:<url> flag, which enables the output of commands to be transmitted to an arbitrary external URL. Additionally, a feedback command is included that can send user-provided data to a remote endpoint if a specific environment variable is configured.
  • [PROMPT_INJECTION]: The skill defines an 'Automatic learning' protocol where the agent is explicitly instructed to execute sequences of commands ('playbooks') and follow 'notes' returned by the recall command. This design creates an indirect prompt injection surface where the tool's data output can direct the agent's future actions and influence its decision-making process.
  • Ingestion points: Data returned by the recall command in the playbook and notes fields.
  • Boundary markers: None identified; the instructions mandate following the tool's output verbatim.
  • Capability inventory: The skill has access to shell execution via the Bash tool.
  • Sanitization: No explicit sanitization or validation of the retrieved playbooks is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 19, 2026, 04:04 PM
Security Audit — agent-trust-hub — pp-prediction-goat