The Agent Skills Directory

[DATA_EXPOSURE]: The skill is designed to scan for sensitive data (tokens, secrets, private keys) within Cursor rule files. However, it implements a robust redaction mechanism in the scripts/scan-cursor-context-risks.mjs file, ensuring that sensitive values are not displayed in the agent's output. The instructions emphasize recommending removal and rotation of found credentials rather than exfiltrating them.
[COMMAND_EXECUTION]: The skill utilizes local Node.js scripts (inventory-cursor-context.mjs, scan-cursor-context-risks.mjs, and backup-cursor-context.mjs) to perform its tasks. These scripts use only standard Node.js libraries (node:fs, node:path), do not perform network operations, and are executed via the command line with explicitly scoped paths.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository files (e.g., .cursor/rules, AGENTS.md). While this presents an indirect prompt injection surface, the skill mitigates this by requiring a structured review report, creating a backup before any changes, and mandating explicit user approval before modifying any files. This human-in-the-loop workflow prevents autonomous execution of malicious instructions found in the data.
[SAFE_PRACTICES]: The skill implements strong safety standards, including a dedicated references/safe-editing-procedure.md and references/conflict-resolution.md. It explicitly forbids silent deletions or modifications and uses timestamped backups to ensure recoverability.

cursor-memory-curator