pp-prediction-goat
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: Installation instructions use
npxandgo installto fetch and execute code from repositories and packages managed by the skill's author.- [DATA_EXFILTRATION]: The CLI tool includes a--deliver webhook:<url>flag, allowing command results to be transmitted to arbitrary external servers via HTTP POST. Additionally, an optional feedback mechanism can be configured to send data to a remote endpoint.- [DYNAMIC_EXECUTION]: The skill implements a 'learning' system that stores command sequences ('playbooks') in a local SQLite database and replays them with variable substitution, which represents a complex dynamic execution surface.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Polymarket and Kalshi, including market titles and comments, which could contain instructions targeting the agent. - Ingestion points: Data retrieved from external prediction market APIs via
prediction-goat-pp-clicommands. - Boundary markers: No explicit separators or 'ignore instructions' warnings are provided for the agent when processing tool output.
- Capability inventory: The skill can execute shell commands, read/write files, and perform network requests via the CLI.
- Sanitization: No sanitization or validation of the fetched API content is described.- [COMMAND_EXECUTION]: The skill operates primarily by calling a feature-rich binary that performs extensive file system, network, and system interactions.
Audit Metadata