The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external data from local files and user descriptions to generate file modifications, which creates a surface for indirect prompt injection. This risk is mitigated by the explicit human-in-the-loop verification step required before any changes are written to disk.
Ingestion points: user-provided change_description and existing collection files.
Boundary markers: None specified to separate untrusted data from instructions.
Capability inventory: File system reading, file writing, and shell command execution.
Sanitization: No explicit sanitization or validation of the input data is mentioned.
[COMMAND_EXECUTION]: Executes a find command to display the resulting file tree after updates are applied.
[SAFE]: The skill implements a robust safety control by requiring the agent to display a unified diff and wait for explicit user confirmation before any file modification occurs.

update-collection