The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local shell script run-checks.sh using bash. This script is located within the plugin's own directory structure as defined by the ${CLAUDE_PLUGIN_ROOT} environment variable.
[INDIRECT_PROMPT_INJECTION]: The skill processes external files located in a user-provided directory (<plugin-dir>). Content from these files is ingested by validation tools like claude plugin validate and check.py, and the resulting output is returned to the agent's context. This creates a surface where malicious instructions embedded in the target plugin files could attempt to influence the agent. The skill includes boundary instructions to mitigate this by directing the agent to output results verbatim and avoid summarization. Ingestion occurs via the plugin files, and capabilities include shell script execution. No specific sanitization of input directory contents is detailed.

check