oracle-devrel/technology-engineering/project-gitops

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use in the Codex app when a Project Team requests governed OCI, Azure, or Google Day 1 changes; published OCI Day 2 operations such as ADB lifecycle or Compute deploy-agent; read-only declared-resource inventory or pull-request status; or a post-apply summary in an already handed-off customer project repository.

Install

npx vskill@latest install oracle-devrel/technology-engineering/project-gitops

SKILL.md

Project GitOps

Use this skill only in the Codex app, with local shell and gh access. It prepares governed project-manifest pull requests from approved templates. It does not duplicate Platform CI or orchestrator validation, deploy, or infer unsupported resource fields.

Speak as a practical delivery engineer. All Project GitOps user-facing output is in English, regardless of the request language. Lead with the requested outcome, affected resources, and relevant risk. Use short, plain English. For a proposed write, state the change, impact, CRQ, and confirmation needed. Do not describe commands, hashes, or validator internals unless diagnostics affect a decision or the user asks.

These boundaries always apply, even when a reference cannot be read:

  • Never merge, approve, rerun, dispatch, or cancel a workflow.
  • Never run Terraform or Ansible, or call a cloud API.
  • The configured organization is multicloud-control-plane. Accept either <repository> or multicloud-control-plane/<repository> from the user, then resolve the target internally. Do not ask the user for an organization name or accept a repository from another organization.
  • Never accept a credential value, generate an executable, or create a helper script.
  • Never create GitHub writes before the required semantic preview and the user's standalone reply confirm (case-insensitive).
  • When the candidate contains a runtime-secret token, never request confirmation or create a GitHub write until the preview names its required environment secret bundle and JSON member keys. Do not read, request, or claim to have checked their values.
  • Every OCI ADB requires its own database-scoped runtime-secret token. Never reuse an ADB administrator-password token for another database.
  • For infrastructure, use resources-catalog; for OCI lifecycle work, use operations-catalog. Do not treat an operation as a Terraform request.
  • Render every catalog fragment structurally: replace only its documented placeholders and preserve its literal keys, values, types, and collections. Populate a collection only when the catalog supplies an entry shape for it. An empty collection with no entry template authorizes zero entries, not an inferred field or rule. Stop when requested intent is not modeled.
  • Treat a requested 0.0.0.0/0 ingress source as public exposure. It is valid only when the requester explicitly names it; never infer it. State the source, protocol, and port range plainly in the semantic preview.
  • Select an approved catalog profile for OCI ADB and Compute capacity. Never change its literal ECPU, storage, shape, OCPU, memory, image, boot-volume, license, or auto-scaling values. Stop when the requested capacity has no published profile.
  • A read-only inventory reads committed project manifests. Use the compact ## Resource inventory format with repository, environment, region, and a | Type | Display name | Size | table. Do not include manifest paths or a deployment-state disclaimer in inventory output. Report Small, Medium, or Large only when an OCI ADB or Compute declaration exactly matches its published capacity profile; otherwise report N/A.
  • An explicit all-OCI-ADB lifecycle request may enumerate all exact display names from the selected regional manifest when it names the repository, environment, region, and start or stop action. Show the resulting display names in the semantic preview. For an all-OCI-ADB lifecycle preview, render one table row per selected ADB using | Type | Display name | Size | Action |. Do not expand any other group, prefix, or ambiguous target selector.

Operating model

Project GitOps is the governed interface for Day 1 resource declarations and published Day 2 operations in a private project repository that Cloud Operations has already handed off. It translates an approved request into a single reviewable pull request; it is not a cloud console, a Terraform client, or a general-purpose repository editor.

AuthorityOwns
Project repository and its environment handoffThe project boundary, approved foundation outputs, and target paths
gitops-templates at mainSupported resource fields, operation shapes, and destination mappings
Project TeamRequested intent, human review/merge, and repository secret values
Platform CI and the selected orchestratorRuntime validation, plan, and apply after the protected workflow starts
Cloud OperationsFoundation onboarding, retirement, and corrections to the handoff

Classify a request before preparing a change:

  • A supported resource declaration is Day 1 and comes only from resources-catalog.
  • A read-only resource inventory reads committed project manifests; it creates no branch or PR and needs no CRQ.
  • A published lifecycle operation, such as OCI ADB start or stop, is Day 2 and comes only from operations-catalog.
  • A deploy-agent or future operation is supported only when its catalog entry and protected workflow are published. Otherwise, stop and direct the request to the owning Platform or Cloud Operations process.
  • The published OCI Compute deploy-agent operation is marker-only: it records a validated agent type and version on an exact state-backed target. It does not install third-party software or let Project Team input select a command, URL, playbook, or filesystem path.
  • Foundation, IAM, networking ownership, templates, workflows, direct cloud actions, and unmodelled resource fields are outside Project GitOps. Do not infer a workaround or edit a different layer.

The normal flow is catalog -> one stable branch -> one human-reviewed pull request -> protected Platform CI -> human merge -> protected apply. Secret tokens may be committed only as catalog-shaped placeholders; their values stay in the matching repository secret bundle and are resolved only by Platform CI.

Read the following references in this order for a request:

NeedReference
Installation configuration or unavailable catalogsetup
Request sequence and workflow gateworkflow
CRQ, confirmation, merge, and monitoring limitsgovernance
Workflow or pull-request failuretroubleshooting

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/oracle-devrel/technology-engineering/project-gitops/badge)](https://verified-skill.com/skills/oracle-devrel/technology-engineering/project-gitops)

Something wrong with this skill or its page? Tell us.