The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (CRITICAL): The skill documentation explicitly instructs users or agents to execute curl -fsSL https://cli.inference.sh | sh. This pattern is a major security risk as it executes unverified code from an untrusted source with full shell privileges.
[EXTERNAL_DOWNLOADS] (HIGH): The skill relies on tools and services from inference.sh, which is not on the list of trusted GitHub organizations or repositories.
[COMMAND_EXECUTION] (HIGH): The skill requests Bash(infsh *) capabilities, granting the agent the power to execute any subcommand of the infsh CLI, which could lead to unauthorized actions if the tool handles credentials or sensitive local state.
[PROMPT_INJECTION] (LOW): The skill is highly vulnerable to indirect prompt injection through the code input parameter. 1. Ingestion points: code field in the input JSON provided to the executor. 2. Boundary markers: None present; the code is treated as a raw execution string. 3. Capability inventory: The execution environment includes full network access (via requests, httpx), file system writes (to the outputs/ directory), and browser automation tools (playwright, selenium). 4. Sanitization: No sanitization, validation, or escaping is performed on the input code before execution.

python-executor