Sealos S3

Sealos S3

Identity and Discovery

• Owner: sealos-s3 (/sealos-s3 and bucket, object, policy, credentials, presign, or quota requests).
• Class: cloud-local-mutation through sealos-cli s3, with an optional redacted deployment handoff.
• Canaries: S3-PRIVATE-REUSE, S3-CONFIRM-PUBLIC, and S3-REDACT-OBJECT.

Scope and Boundaries

Accept a project path and object-storage intent. Analyze first, list existing buckets, create or reuse private storage, initialize credentials only when needed, and wire the smallest existing env-key set. Preserve local MinIO/Compose fallback; public policy, rotation, deletion, and destructive object changes remain gated.

Risk and Confirmation

Keep buckets private by default. Ask before public policy, credential rotation for an active app, bucket/object deletion, or replacing storage configuration. Never print secret keys, full credential blocks, kubeconfig, auth files, or copied env values; parse JSON output rather than scraping secrets from tables.

Lifecycle Workflow

For each request, resolve the project, analyze storage need, confirm CLI/auth/region/workspace, list before create/reuse, wait for credential readiness, wire existing keys, prove an authenticated upload/read or presign path, clean temporary objects, and report policy state. Emit request-scoped success, stopped, or error; the existing private-first workflow remains the domain extension below. Workspace ambiguity, missing credential readiness, or a tracked env target stops the request before mutation.

Request Contract

input:
  project: local path or repository source
  intent: bucket purpose, object path, and sharing requirement
  policy: private by default; publicRead/publicReadwrite only after confirmation
preconditions:
  - analyzer evidence identifies the storage adapter and env keys
  - sealos-cli/auth/region/workspace are resolved
  - bucket list and credential status are available

The ordered action is analyze -> resolve account/workspace -> list -> create or reuse -> wait for secret -> wire -> object-flow -> cleanup. Bucket identity, policy, namespace, and env mutation remain request-scoped.

Progressive Disclosure

Load analyzer, bucket, policy, credential, object-flow, and cleanup procedures one level deep when their phase is reached. Keep private policy, confirmation, and redaction canaries visible before the CLI detail.

Output, Stop, and Error States

• success: bucket identity/policy/readiness, workspace, env key names, authenticated object or presign proof, cleanup state, and redaction status.
• stopped: ambiguous workspace, unavailable credential readiness, tracked env target, public policy, rotation, replacement, or destructive confirmation boundary with the safe private/presign alternative; no gated operation is claimed.
• error: analyzer, CLI, policy, credential, object, or env-write step, sanitized diagnostic category, affected artifact, and recovery action with credentials, endpoints, and connection values redacted.

Handoffs

An optional deployment handoff uses the complete tuple below. The receiver re-checks deployment scope and runtime evidence.

target: sealos-deploy
inputArtifact: private bucket identity, policy, approved env-key contract, credential-readiness state, and object-flow proof
allowedAction: consume approved ObjectStorageBucket/Secret wiring within the selected deployment scope
failureReturn: sanitized analyzer, CLI, policy, credential, object, or env diagnostic with the failed phase
responseOwner: sealos-s3

Direct S3 requests use target: none and keep the same evidence fields.

Verification

Use analyze-project-s3.mjs, sealos-cli s3 JSON output, authenticated object-flow evidence, and baseline cases s3-positive-private-round-trip and s3-violating-unconfirmed-public-or-rotation. Verify private policy, env preservation, and redaction before success.

Use this skill to give a project real Sealos object storage through sealos-cli s3. The default outcome is: identify the app's object-storage need, create or reuse a bucket, initialize credentials only when needed, wire the smallest safe set of local env vars, and verify the project's upload/download or presigned URL path.

This skill is grounded in zjy365/sealos-cli#28, which registered the s3 command and implemented bucket CRD operations plus S3-compatible object operations.

Safety Rules

1. Never print secret keys, full S3 credential blocks, or copied env values in the final answer.
2. Do not overwrite an existing env value without confirming or preserving the old value.
3. Do not commit .env, .env.local, S3 access keys, secret keys, kubeconfig, or Sealos auth files.
4. Ask before making a bucket public. Default bucket policy is private.
5. Ask before destructive operations: s3 delete-bucket, s3 delete, credential rotation for an active app, or replacing app storage configuration.
6. Use JSON output from sealos-cli by default and parse it instead of scraping table output.
7. Treat s3 secret output as sensitive even though the CLI can print it.

Workflow

1. Resolve the target project

Confirm the working directory with pwd or git rev-parse --show-toplevel.

Run the analyzer when a project directory is available:

node <SKILL_DIR>/scripts/analyze-project-s3.mjs <project-dir>

Use the analyzer result as a starting point, then inspect the real files it cites before editing anything. It intentionally avoids printing secret values.

2. Check sealos-cli

Prefer an existing sealos-cli binary:

sealos-cli --version
sealos-cli s3 --help
sealos-cli whoami

If it is not installed, use npx -y sealos-cli@latest ... for one-off commands. Ask before installing it globally.

If auth is missing or expired, run:

sealos-cli login <region>
sealos-cli workspace list
sealos-cli workspace current

Use the workspace the user expects. If multiple workspaces exist and the target is ambiguous, ask before provisioning. sealos-cli s3 derives the object-storage user from the active kubeconfig namespace, so a wrong workspace means wrong buckets and credentials.

3. Choose create or reuse

List existing buckets first:

sealos-cli s3 buckets -o json

Reuse an existing bucket when its purpose and policy match. Create a new one when the project has no suitable bucket or the user asks for a fresh bucket:

sealos-cli s3 create-bucket <bucket-name> --policy private -o json

Use private unless the user explicitly needs public reads or writes. Bucket policies accepted by the PR are private, publicRead, and publicReadwrite; aliases such as public-read normalize to publicRead, but use canonical values in instructions and scripts.

4. Initialize credentials only when needed

For app env wiring or object operations, fetch credentials:

sealos-cli s3 secret -o json

The command creates the ObjectStorageUser if it does not exist, then waits briefly for status. If credentials are not ready, retry after a few seconds instead of creating raw CRDs by hand.

Use references/sealos-cli-s3.md for the current command contract and response handling.

5. Wire the development environment

Map only the keys the project already uses. Common targets:

Use endpoint from secret.external for local laptop development. Use secret.internal only when the app runs inside Sealos/Devbox and the runtime can reach the internal endpoint.