The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local utility script $CLAUDE_PLUGIN_ROOT/scripts/file_metrics.py to determine file characteristics. This is a local script within the skill environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data via the Read tool, creating a surface for indirect prompt injection. It implements mitigation via a 'quote-grounding' technique (extracting verbatim text before summarizing) and enforces a strict structured summary template, which helps prevent the model from obeying instructions embedded in the summarized files.
[CREDENTIALS_UNSAFE]: The instructions explicitly direct the agent to identify the presence of credentials in configuration files without exposing the actual sensitive values in the output summary.

file-summarization