Azure Python App Service Deployer

The skill narrowly scopes itself to Azure App Service Linux and explicitly hands off networking, Key Vault, databases, and IaC. It shows derived resource defaults before creation and excludes .env, virtual environments, and caches. However, it defaults to a paid P0v3 plan, explicitly proceeds without confirmation, and does not explain cost, permission scope, sensitive-data flow, or rollback, so substantial points are deducted.

The main workflow, idempotent show-or-create pattern, transient retry policy, and error matrix are reasonably clear. However, framework detection is largely dependency-token based, FastAPI defaults to main:app even when the real entry point may differ, and the creation reference requires verification while the main rules prohibit post-deployment verification. These inconsistencies and the lack of execution evidence require a conservative deduction.

Trigger phrases, supported frameworks, exclusions, and the azure.yaml branch are explicit, with Bash and PowerShell coverage. Unknown frameworks are deployed despite likely startup failure, and entry-point handling is limited. The core function depends on Azure CLI, azd, Azure App Service, and overseas cloud access, with no Chinese-language or mainland-China reachability guidance, so points are deducted.

The skill is well layered across SKILL.md and focused reference files for creation, detection, deployment, errors, startup, and retries. MIT licensing, author, and version metadata are present, while repository README, SECURITY, and CI provide some governance signals. It lacks a skill-specific changelog, dependency/version matrix, installation notes, cost and permission limitations, and explicit maintenance ownership, so it does not merit full marks.

The documented flow covers resource creation, framework detection, packaging, deployment, and endpoint reporting, so it plausibly completes the code-push task. FastAPI entry-point assumptions, unknown-framework startup gaps, asynchronous build behavior, and the prohibition on post-deploy verification limit directly usable completeness. Static calibration also prevents awarding more than 7, so the score is conservative.

The references provide concrete commands, retry rules, an error matrix, and repository-level CI/test evidence. The supplied tests do not cover this skill's actual Azure deployment paths, and there is no skill-specific acceptance suite or third-party execution record. The assessment therefore rests mainly on static text and receives only limited credit.