The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Attack Surface: The skill is designed to analyze and improve user-specified target artifacts (files, folders, or documents), which acts as an ingestion point for untrusted data.
Ingestion points: Target artifacts are read into the agent's context using Read, Grep, and Glob tools as defined in SKILL.md and SKILL.ko.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to isolate untrusted content from the agent's core logic during its multi-pass reasoning process.
Capability inventory: The skill possesses powerful capabilities, including the ability to modify local files (Edit, Write) and execute shell commands via the Bash tool for validation.
Sanitization: There is no mention of sanitizing or validating the contents of the target files before they are processed by the reasoning logic or used in shell commands.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform validation tasks such as running tests, linters, and build commands. While these are legitimate developer tasks, they represent a vector for code execution that could be exploited if the agent's behavior is subverted by malicious content placed within the target files being improved.

improve