The Agent Skills Directory

[COMMAND_EXECUTION]: The process instructions in SKILL.md (Steps 3 and 4) direct the agent to construct and execute shell commands using string interpolation of user-controlled variables like {path}, {content}, and {label}. By placing these variables directly into a shell command string, the skill creates a high risk of command injection if the input contains shell metacharacters such as semicolons, backticks, or command substitution sequences.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a significant attack surface by processing untrusted data from user-provided files and text blocks and passing them into shell-based tool executions.
Ingestion points: User-specified file paths, directory contents (Step 2), and inline content blocks.
Boundary markers: There are no markers or delimiters used to ensure that the content being processed is not interpreted as part of the shell command structure.
Capability inventory: The skill possesses the capability to execute shell commands via the local scripts eval-runner.py and quality-tracker.py.
Sanitization: The instructions lack any requirement or logic for sanitizing or escaping the user-provided strings before they are interpolated into the shell command arguments.

eval-suite