The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The invoke action allows for the execution of arbitrary Python code within a remote sandbox environment managed by the vendor infrastructure at api.agentpmt.com.
[EXTERNAL_DOWNLOADS]: The skill provides installation instructions that use npx skills add to fetch and configure auxiliary setup skills from the vendor's repository at github.com/AgentPMT/agent-skills.
[DATA_EXFILTRATION]: The sandbox environment includes the requests library, which enables the execution of network operations to external domains as part of the tool's intended API integration and web scraping capabilities.
[PROMPT_INJECTION]: The skill processes code provided via the code parameter, which presents a surface for indirect prompt injection if the input is derived from untrusted sources.
Ingestion points: The code parameter within the invoke action defined in schema.md.
Boundary markers: No explicit delimiters or instructions for the agent to ignore embedded commands within the input code are defined.
Capability inventory: The environment supports arbitrary Python code execution, network access via requests, and file/data manipulation using numpy and pandas as described in SKILL.md.
Sanitization: The skill documentation does not describe explicit sanitization or validation of the input code string before execution.

secure-python-code-sandbox