Azure Identity for Python

The documentation promotes Entra authentication, managed identity, environment-based secrets, and credential context managers, and warns against hardcoded credentials and unencrypted cache fallback. However, DefaultAzureCredential attempts multiple local credential sources by default, debug logging may expose sensitive authentication diagnostics, and the wildcard tenant example permits any tenant. Least-privilege, confirmation, data-flow, and recovery boundaries are not fully specified, so points are deducted.

The main and reference documents are broadly consistent on sync, async, caching, error handling, and logging, and they distinguish credential-unavailable from authentication-failure errors. However, the Python package version is not pinned, current chain order and parameters depend on external documentation, and no executable tests for this skill's key paths are provided, so the static score is conservative.

Triggers, audience, and common environments are reasonably clear, covering local development, Azure hosting, CI/CD, desktop, and headless use. Non-fit scenarios, permission prerequisites, input/output boundaries, and false-trigger conditions are not explicit; Chinese-language guidance is absent, with mainland-China support limited to sovereign-cloud examples, so points are deducted.

The skill has front matter, MIT licensing, author and version metadata, installation notes, layered reference files, an environment matrix, and external API references. It lacks a skill-specific changelog, dependency-version policy, explicit maintenance/update path, FAQ, and complete limitations disclosure; some reference material is duplicated, so points are deducted.

Coverage includes DefaultAzureCredential, managed identity, service principals, workload identity, token providers, async usage, and error handling, giving directly useful guidance for common authentication tasks. Most examples use placeholders, and there is no demonstrated end-to-end validation, permission setup, or comparison with alternatives. Static review cannot confirm direct execution against the target package version, so the score is capped at the static limit.

The material names traceable primary sources such as Microsoft Learn, PyPI, and the Azure SDK source, while repository context shows generic CI, evaluation workflows, and test infrastructure. However, the supplied files do not demonstrate skill-specific testing or independent reproduction, and the references are not pinned to a precise dependency version, so the score stays at the static ceiling.