The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests market data (questions, descriptions) from external Polymarket APIs and displays them to the user or agent, which could contain malicious instructions.
Ingestion points: scripts/paper_engine.py and scripts/health_check.py fetch data from gamma-api.polymarket.com and clob.polymarket.com.
Boundary markers: None present; the skill displays market titles and questions without delimiters or 'ignore' instructions.
Capability inventory: The agent is instructed to execute various Python scripts (paper_engine.py, execute_paper.py, etc.) which perform file writes to ~/.polymarket-paper/portfolio.db and network requests.
Sanitization: Standard JSON parsing is used, but there is no sanitization or filtering to detect or prevent embedded prompt injection patterns in the market data.
[COMMAND_EXECUTION]: The documentation includes a command pattern that pipes remote data directly into an interpreter.
Evidence: SKILL.md provides an example command: curl -s 'https://gamma-api.polymarket.com/markets...' | python3 -c "...".
Details: While the Python code itself is provided as a static string literal, the pattern of piping unverified remote data into a language interpreter carries inherent risk if the processing logic is flawed or the remote data is malicious.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch real-time market data.
Evidence: Requests are made to gamma-api.polymarket.com and clob.polymarket.com throughout the scripts.
Details: These requests target official endpoints for Polymarket services and are necessary for the skill's primary purpose of live price simulation.

polymarket-paper-trader