The Agent Skills Directory

Command Execution from External Files: The skill is designed to parse and execute shell commands extracted from SKILL.md files. While this is necessary for its purpose as a validation utility, it creates a situation where the agent executes commands found in text files, which could lead to unintended operations if the source files are not verified.
Indirect Prompt Injection Surface: The process of reading external documentation content and using it to perform shell actions creates a surface for indirect injection. If a source file contains adversarial shell commands, this skill would execute them as part of its validation routine.
Ingestion points: The skill reads SKILL.md files to identify executable command blocks (Step 2).
Boundary markers: Absent; the instructions do not define specific delimiters or security boundaries for the extracted content beyond OS matching.
Capability inventory: The skill relies on full shell execution (bash, powershell) to verify documented commands (Step 4).
Sanitization: Absent; there is no documented validation or filtering logic to ensure the commands are safe before they are run.
Dynamic Execution Pattern: The skill performs runtime adaptation of commands, including placeholder substitution and cross-platform translation. This dynamic handling is part of its functional requirements for testing but involves executing code paths derived from external text data.

validate-skills