Lists multiple specific concrete actions: 'runs azd up, azd deploy, terraform apply, and az deployment commands with built-in error recovery.' Also specifies prerequisites like 'infra/{iac}/{project}/.azure/plan.md' and 'validated status from azure-validate.'

Clearly answers both 'what' (execute Azure deployments for already-prepared applications, runs specific commands with error recovery) and 'when' (explicit WHEN clause with trigger phrases). Also includes explicit 'DO NOT USE WHEN' guidance to prevent misuse, which goes above and beyond.

Excellent coverage of natural trigger terms users would say: 'run azd up', 'push to production', 'push to cloud', 'go live', 'ship it', 'bicep deploy', 'terraform apply', 'publish to Azure', 'launch on Azure'. These are highly natural phrases a user would actually type.

Exceptionally distinctive — explicitly differentiates itself from 'azure-prepare' with clear boundary conditions. The 'DO NOT use this skill when...' and 'DO NOT USE WHEN' clauses with specific counter-examples ('create and deploy', 'build and deploy') make it very unlikely to conflict with the preparation skill.

The content is extremely verbose and repetitive. The prerequisite check about plan status being 'Validated' and not manually updating it is stated at least 4-5 times across the prerequisite block, rules, steps, gotchas, and validation proof check. The triggers section explains scope boundaries that are already covered in the description. Multiple warning blocks repeat the same information.

The steps table provides a clear sequence but lacks concrete executable commands — it delegates almost everything to external references (recipes, checklists, verification docs). There are no inline code examples showing actual azd/terraform commands. The MCP tools table is helpful but the actual deployment execution is entirely deferred to recipe files.

The workflow is clearly sequenced with explicit prerequisite gates (steps 0-1), a pre-deploy checklist (step 2), execution (steps 3-4), post-deploy (step 5), error handling (step 6), and verification (step 7). Validation checkpoints and feedback loops are well-defined, including the auto-prepare gate and validation proof checks.

Content is well-structured with clear one-level-deep references. The Reference Index table with 'When to Load' guidance is excellent for on-demand loading. SDK references, recipes, and troubleshooting are all clearly signaled with direct links. The instruction to 'Load these on demand — do NOT read all at once' is appropriate.