The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides installation instructions (curl -fsSL https://cli.inference.sh | sh) that pipe a remote script directly into the shell. This pattern is highly risky as the contents of the script are executed immediately without verification or user review.
[EXTERNAL_DOWNLOADS]: The skill relies on fetching an executable binary and an installation script from external domains (cli.inference.sh and dist.inference.sh) which are not recognized as trusted sources.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute infsh commands. While the allowed-tools frontmatter restricts the command scope, the execution of arbitrary code via these commands is the core functionality.
[DATA_EXFILTRATION]: By design, this skill sends Python code and potentially local data to the inference.sh remote API for execution. This creates a surface for exfiltrating sensitive environment variables or local files if the agent is directed to do so.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8).
Ingestion points: The code field in the input schema for infsh app run in SKILL.md accepts arbitrary strings.
Boundary markers: None. There are no instructions to the agent to validate or sanitize the code before execution.
Capability inventory: The skill facilitates arbitrary Python execution including networking (requests, httpx) and file manipulation through the infsh tool.
Sanitization: Absent. The skill passes the raw input directly to the external execution environment.

python-executor