AGENTS.md
+18 −7
https://raw.githubusercontent.com/swadeshitech/prodhub-agent-kit/HEAD/AGENTS.md
− and [failure recovery](docs/failure-recovery.md) for supporting detail.− 3. For Jenkins builds, call `get_prodhub_build_pipeline_contract` before deriving profile values.− 4. Inspect the checked-out repository. Use repository evidence, not guessed ports,− 5. For new applications, propose onboarding before creating records. The− 6. Proceed within the user's authorized scope. Ask for approval when it is− 7. After approval, call `provision_application_preview`, then poll− 8. On failure, call `summarize_pipeline_failure` when the tool is available and+ [runtime templates](docs/runtime-templates.md), and [failure recovery](docs/failure-recovery.md)+ for supporting detail.+ For Kubernetes capacity management, read [the Resource Manager runbook](docs/resource-manager.md).+ 3. For Resource Manager requests, call `get_resource_manager_contract`, then+ `list_product_operations` with `product_area_link` set to+ `/dashboard/resourceManager`. Resolve current groups and rules before proposing+ changes. Require explicit approval before create, update, or delete. Cloud+ Resources inventory is not exposed through MCP.+ 4. For Jenkins builds, call `get_prodhub_build_pipeline_contract` before deriving profile values.+ 5. Inspect the checked-out repository. Use repository evidence, not guessed ports,+ 6. Before creating or updating a runtime template, call+ `get_runtime_template_contract`. For a repository-hosted Helm chart, verify+ `Chart.yaml` and use a repository-relative chart path. Inherit application+ SCM/repository/commit unless the user explicitly requests an override.+ 7. For new applications, propose onboarding before creating records. The+ 8. Proceed within the user's authorized scope. Ask for approval when it is+ 9. After approval, call `provision_application_preview`, then poll+ 10. On failure, call `summarize_pipeline_failure` when the tool is available and