The Agent Skills Directory

Credential Management and Storage: The skill manages Mural OAuth 2.0 credentials (access/refresh tokens and client secrets). It prioritizes storage in the OS keyring and falls back to local configuration files with strict 0600 permissions. It includes a mechanism to promote file-based credentials to the keyring automatically.
Automated Data Redaction: A robust redaction engine is implemented to scrub tokens, secrets, and sensitive query parameters from log output and error messages, providing defense-in-depth against accidental information disclosure.
Dynamic Module Loading for Backends: The environment variable MURAL_KEYRING_BACKEND allows for the dynamic instantiation of custom keyring implementations via importlib.import_module. While this is an intended extension point for authentication, it represents a potential area for review as it allows loading arbitrary Python classes if the environment is compromised.
Network and SSRF Protection: Network requests to Mural and Azure Blob Storage are performed through dedicated openers that refuse HTTP redirects. Asset upload URLs are validated against a strict host allowlist to mitigate potential SSRF vulnerabilities.
Indirect Prompt Injection Surface: The skill retrieves user-authored content from Mural widgets (sticky notes, textboxes). The documentation explicitly identifies this as untrusted content that downstream agents must handle carefully, as the CLI itself does not perform content sanitization beyond JSON encoding.
Capped Data Ingestion: To prevent potential denial-of-service via resource exhaustion, the skill uses capped readers for HTTP response bodies and enforces size limits on pagination cursors.

mural