The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation includes a command to install the CLI tool using curl -fsSL https://cli.inference.sh | sh. This pattern is highly risky as it executes a remote shell script without verification, potentially allowing for arbitrary code execution from an external source.\n- [DYNAMIC_EXECUTION]: The skill provides Python code snippets for generating charts (using matplotlib) and HTML templates for images, both of which are executed at runtime via the infsh tool. This creates a surface for dynamic code execution based on instructions provided to the agent.\n- [EXTERNAL_DOWNLOADS]: The skill uses npx skills add to fetch and execute additional agent skills from remote repositories, which introduces a dependency on the integrity of the NPM registry and the third-party skill sources.\n- [COMMAND_EXECUTION]: The skill is configured with allowed-tools: Bash(infsh *), which gives the agent broad authority to execute the infsh CLI tool with any arguments, including those that generate and save files to the local system.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data to generate visual content through HTML and Python templates without explicit sanitization or boundary markers. 1. Ingestion points: Command inputs in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Subprocess calls via Bash, file writing via Python (matplotlib). 4. Sanitization: Absent.

pitch-deck-visuals