The Agent Skills Directory

[COMMAND_EXECUTION]: The documentation includes examples of potentially destructive shell commands (e.g., rm -rf ~/.cursor/index/) intended for manual maintenance and clearing local application data.
[DATA_EXFILTRATION]: The skill instructions and configuration examples reference numerous sensitive file paths and patterns (e.g., .env files, SSH private keys like id_rsa, credentials.json, and cloud provider tokens) as targets for exclusion from the AI context. While these are listed for protection purposes, the agent is instructed to identify and handle these paths during the configuration process.
[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to inspect and process workspace files to enforce privacy settings, which constitutes an attack surface for instructions embedded in the repository data.
Ingestion points: Local workspace files, repository configuration, and sensitive directory structures during the privacy boundary setup.
Boundary markers: Absent. The skill does not provide specific instructions to distinguish between configuration metadata and potentially malicious instructions embedded in the files being processed.
Capability inventory: The skill is configured with access to Read, Write, Edit, and Bash(cmd:*) tools, providing a wide range of actions the agent could take if influenced by external content.
Sanitization: There are no specified procedures for sanitizing or validating workspace data before it is processed by the agent.

cursor-privacy-settings