The Agent Skills Directory

[SAFE]: No malicious patterns or high-risk vulnerabilities were identified. The skill provides integration logic for CryptPad with a focus on security boundaries and secret management.\n- [COMMAND_EXECUTION]: The scripts/test_cryptpad.py file executes local shell commands using subprocess.run to perform unit tests. Specifically, it runs the cryptpad.py script and a Node.js process to verify callback logic. These are standard development practices and do not expose the host system to external command injection.\n- [EXTERNAL_DOWNLOADS]: The probe tool in scripts/cryptpad.py makes network requests to user-provided origins to discover CryptPad instance settings. It implements safety features including origin validation (restricting non-HTTPS to loopback), refusal of HTTP redirects, and a 1MiB response size limit.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from instance probes and document links. Ingestion points: External configuration endpoints and stdin link input. Boundary markers: Explicit instructions to confirm actions, verify state via browser, and report boundary failures. Capability inventory: Limited to read-only network discovery and file-link parsing. Sanitization: Origin validation, regex-based JSON extraction to avoid JS execution, and redaction of query/fragment secrets.

cryptpad