The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The primary purpose of the skill is to execute arbitrary Python code on a remote infrastructure via the infsh CLI. The code is sent as a JSON payload to the inference.sh service.
[COMMAND_EXECUTION]: The skill requires access to the Bash tool to run the infsh command-line utility for logging in and running applications.
[EXTERNAL_DOWNLOADS]: The documentation points to external installation scripts for the CLI hosted on GitHub (raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md).
[PROMPT_INJECTION]: The skill facilitates indirect prompt injection risks through its web scraping and data processing capabilities.
Ingestion points: The skill explicitly supports fetching untrusted data from external websites using pre-installed libraries like requests, BeautifulSoup, Selenium, and Playwright (e.g., in SKILL.md examples).
Boundary markers: There are no instructions or delimiters provided to help the agent distinguish between the execution results and potential malicious instructions embedded in scraped content.
Capability inventory: The skill possesses high-privilege capabilities including network access, file system writes (to the outputs/ directory), and arbitrary code execution.
Sanitization: The skill does not implement or require sanitization of the data retrieved from the web before it is incorporated into the agent's context.

python-executor