[codex] Add Crownest provider by tristanmanchester · Pull Request #681 · openclaw/crabbox

tristanmanchester · 2026-06-25T04:23:36Z

Summary

Adds a Crownest delegated-run provider with archive sync, Workspace Run execution, provider docs, matrix/category updates, config/env handling, and a live smoke script.
Hardens Crownest lifecycle behavior around claim cleanup, failed setup, cancellation, kept sandboxes, status waits, cleanup command scoping, stream failures, per-lease operation locking, and redaction of presigned upload errors.
Preserves and cleans sandbox IDs from malformed partial create responses, sends the exact archive Content-Length to staged upload targets, and keeps live-smoke reuse and cleanup in the original provider scope.
Adds focused Go coverage for provider/client/CLI behavior plus script coverage for live smoke classification.

Why / impact

This lets Crabbox run workloads on Crownest while keeping the core provider-neutral and avoiding secret persistence or secret forwarding. The provider treats the Crownest API key as runtime-only configuration, does not pass it on the command line, rejects unsupported delegated-run surfaces, and documents the current environment-forwarding limitation.

Review comment addressed

Removed unrelated worker/package-lock.json drift from the branch.
Added Crownest per-lease operation locking for retained run reuse, stop, and cleanup paths so concurrent local processes cannot race over the same retained sandbox.
Made retained-session cleanup commands carry the Crownest selector scope (--crownest-url, --crownest-project-id, --crownest-template) so custom-scoped retained sandboxes can be stopped from the printed command.
Addressed Codex review on one-shot cleanup: when Crownest owns cleanup with keepSandbox=false, Crabbox removes only the local claim and does not call DeleteSandbox; retained sandboxes still use explicit delete when Crabbox requested retention and later cleans up.
Addressed maintainer autoreview findings by making archive uploads non-chunked and carrying the explicit Crownest template through status, reuse, stop, and trap cleanup.

Repo guidance checked

AGENTS.md: provider-neutral core, no generated outputs edited by hand, verification commands included in the PR, and no provider/broker tokens in code/docs.
SECURITY.md: no live credentials or private infrastructure details are published here.
.github/: no pull request template is present; CI target is the repository default branch, main.
Crownest API docs referenced for provider integration context: https://docs.crownest.dev/docs

Validation

go test ./internal/providers/crownest ./internal/cli
Passed.
go test -race ./...
Passed on the final current-main merge.
go vet ./...
Passed.
Filtered core coverage used by CI
90.1% after focused Crownest config-path tests; threshold is 90.0%.
go build -trimpath -o bin/crabbox ./cmd/crabbox
Passed. macOS linker emitted only: ld: warning: ignoring duplicate libraries: '-lobjc'.
scripts/check-docs.sh
Passed: checked 51 command docs, provider matrix with 69 providers, 202 markdown files, and built dist/docs-site.
node --test scripts/live-crownest-smoke.test.js
Passed: 1 test.
Full Worker format/lint/typecheck/test/build gate
Passed: 752 tests and Wrangler dry-run build.
git diff --exit-code origin/main -- worker/package-lock.json
Passed: worker lockfile is no longer part of this PR.
Whole-branch autoreview plus focused post-fix autoreview
Two actionable findings fixed; final review clean with no accepted/actionable findings.

Live Crownest evidence

Using a Crownest API key from the shell environment only, I ran bin/crabbox doctor --provider crownest and a real remote workload against a temporary repo containing six files. The remote command ran Node tests, analyzed a CSV dataset, wrote a summary artifact, and verified the aggregate output.

Key output:

ok      provider provider=crownest timeout=10s auth=ready api=ready mutation=false runtime=ready
sync candidate: 6 files, 8.7 KiB
leased [redacted] slug=[redacted] provider=crownest sandbox=[redacted]
v22.22.3
# pass 3
analysis complete {"totalEvents":240,"totalFailures":34,"totalCostCents":6582}
verification complete {"totalEvents":240,"totalFailures":34,"totalCostCents":6582,"teams":[{"team":"alpha","count":60,"failures":8,"failureRate":0.1333,"p95DurationMs":851,"totalCostCents":1680},{"team":"beta","count":60,"failures":9,"failureRate":0.15,"p95DurationMs":838,"totalCostCents":1606},{"team":"delta","count":60,"failures":8,"failureRate":0.1333,"p95DurationMs":820,"totalCostCents":1628},{"team":"gamma","count":60,"failures":9,"failureRate":0.15,"p95DurationMs":783,"totalCostCents":1668}]}
crownest run summary sync=58ms command=4.898s total=17.073s exit=0

I also ran the live smoke script with CRABBOX_BIN=bin/crabbox scripts/live-crownest-smoke.sh. It exercised doctor, one-shot run, keep, status, reuse, and stop:

released lease=[redacted] sandbox=[redacted]
classification=live_crownest_smoke_passed

Config / secrets

Crownest requires CRABBOX_CROWNEST_API_KEY or CROWNEST_API_KEY to be supplied at runtime.
The API key is intentionally not persisted in Crabbox config and is not passed as a command-line argument.
Provider docs cover the current delegated-run behavior and note that command environment forwarding is not supported yet.

clawsweeper · 2026-06-25T04:24:33Z

Codex review: needs changes before merge. Reviewed July 2, 2026, 2:00 PM ET / 18:00 UTC.

Summary
The PR adds a built-in Crownest delegated-run provider with CLI config/env handling, provider registration, docs, tests, and a live smoke script.

Reproducibility: yes. for the review findings by source inspection: client.go still probes /v1/sandboxes?limit=1, and backend.go publishes a claim before acquiring the operation lock. I did not run live Crownest credentials in this read-only review.

Review metrics: 2 noteworthy metrics.

Provider scope: 1 provider added across 23 files. A built-in provider addition changes CLI config, docs, provider registration, runtime behavior, and live smoke coverage rather than a narrow local fix.
Diff size: 3704 added, 3 removed. The amount of new provider code makes lifecycle and permission-boundary review important beyond green CI.

Merge readiness
Overall: 🦐 gold shrimp
Proof: 🦞 diamond lobster
Patch quality: 🦐 gold shrimp
Result: needs maintainer review before merge.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

[P2] Fix the doctor probe to match least-privilege Workspace Run credentials.
Acquire the Crownest operation lock before publishing a newly created local claim.

Risk before merge

[P1] Merging adds Crownest as a built-in third-party provider, so maintainers need to accept ongoing docs, config, lifecycle, credential-scope, and support ownership for the Workspace Run API.
[P1] Normal CI cannot prove Crownest account permission boundaries, quota behavior, staged upload behavior, event streaming, or retained-sandbox lifecycle; the PR body's live proof is strong but does not replace maintainer provider acceptance.

Maintainer options:

Fix the two Crownest blockers (recommended)
Update doctor to validate a supported Workspace Run/auth surface and acquire the per-lease operation lock before publishing any new local claim, with focused regression tests.
Pause provider acceptance
If maintainers are not ready to own Crownest as a built-in provider, keep the PR waiting or close it until the support direction is confirmed.

Copy recommended automerge instruction

@clawsweeper automerge

Special instructions:
Fix Crownest doctor so least-privilege Workspace Run credentials are not rejected by a sandbox-list probe, and fix newly acquired Crownest runs so the per-lease operation lock is held before the local claim/slug is published; add focused regression tests for both paths.

Next step before merge

The remaining blockers are narrow code/test repairs in the Crownest provider rather than an unclear product decision, although final provider acceptance remains maintainer-owned.

Security
Cleared: The diff adds provider HTTP calls, secret env handling, and a live smoke script, but I found no concrete credential persistence, argv secret, or supply-chain regression in the inspected patch.

Review findings

[P2] Probe Workspace Run auth in doctor — internal/providers/crownest/client.go:159-160
[P2] Acquire the operation lock before publishing new claims — internal/providers/crownest/backend.go:216-219

Review details

Best possible solution:

Keep the Crownest implementation behind the provider adapter, but fix the doctor permission boundary and claim/operation-lock ordering before final maintainer provider acceptance.

Do we have a high-confidence way to reproduce the issue?

Yes for the review findings by source inspection: client.go still probes /v1/sandboxes?limit=1, and backend.go publishes a claim before acquiring the operation lock. I did not run live Crownest credentials in this read-only review.

Is this the best way to solve the issue?

No; the provider shape is close, but the current implementation needs the doctor probe and claim/lock ordering fixed before it is the narrow maintainable solution.

Full review comments:

[P2] Probe Workspace Run auth in doctor — internal/providers/crownest/client.go:159-160
doctor still calls Probe, and Probe performs GET /v1/sandboxes?limit=1. A Crownest key that can create/read/cancel Workspace Runs but cannot list sandboxes can run the delegated workload but will fail doctor as an auth/API failure, so the probe should exercise the Workspace Run surface or treat sandbox-list denial as non-fatal.
Confidence: 0.88
[P2] Acquire the operation lock before publishing new claims — internal/providers/crownest/backend.go:216-219
claimAcquiredSandbox writes the local claim and makes the slug resolvable before this process holds the per-lease operation lock. Another crabbox stop --provider crownest <slug> can acquire the lock first, delete the sandbox, and remove the claim while the active run is still uploading/finalizing/starting; hold the lock before making the claim visible.
Confidence: 0.9

Overall correctness: patch is incorrect
Overall confidence: 0.9

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against c0363ba1ba3b.

Label changes

Label changes:

add merge-risk: 🚨 compatibility: The doctor probe can reject valid Crownest run-only credentials if sandbox-list permission is not granted.
add merge-risk: 🚨 availability: A local stop racing a newly published claim can delete the sandbox while the active run is still starting.
remove merge-risk: 🚨 other: Current PR review merge-risk labels are merge-risk: 🚨 compatibility, merge-risk: 🚨 availability.

Label justifications:

P2: This is a normal-priority provider addition with concrete pre-merge correctness blockers but no shipped-user regression.
merge-risk: 🚨 compatibility: The doctor probe can reject valid Crownest run-only credentials if sandbox-list permission is not granted.
merge-risk: 🚨 availability: A local stop racing a newly published claim can delete the sandbox while the active run is still starting.
rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦞 diamond lobster and patch quality is 🦐 gold shrimp.
status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Sufficient (terminal): The PR body includes redacted live terminal output for doctor, a real remote workload, and the live smoke script covering doctor, one-shot run, keep, status, reuse, and stop.
proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes redacted live terminal output for doctor, a real remote workload, and the live smoke script covering doctor, one-shot run, keep, status, reuse, and stop.

Evidence reviewed

Acceptance criteria:

[P1] go test ./internal/providers/crownest ./internal/cli.
[P1] go vet ./...

What I checked:

AGENTS.md policy applied: Read the full target AGENTS.md and applied the provider-neutral adapter boundary and secret-handling guidance to the review. (AGENTS.md:1, c0363ba1ba3b)
Doctor probes sandbox list: Crownest Doctor delegates to api.Probe, and Probe performs GET /v1/sandboxes?limit=1 instead of validating the Workspace Run surface. (internal/providers/crownest/client.go:159, ed1e87e9bedb)
Claim is published before the operation lock: Run calls claimAcquiredSandbox before lockAcquiredLease, so the new slug can become visible before the creating process owns the per-lease lock. (internal/providers/crownest/backend.go:216, ed1e87e9bedb)
Operation-lock tests miss new-claim publication: Existing lock tests cover reuse, stop, cleanup, and lock cancellation, but not the newly acquired run path where the claim is first published. (internal/providers/crownest/backend_test.go:298, ed1e87e9bedb)
Live behavior proof in PR body: The PR body includes redacted terminal output for doctor, a real Crownest remote workload, and a live smoke script covering doctor, one-shot run, keep, status, reuse, and stop. (ed1e87e9bedb)
Area history sampled: Current main and branch history show recent provider catalog, provider mutation, and Crownest hardening work by Peter Steinberger plus adjacent built-in provider addition work by Momo. (c0363ba1ba3b)

Likely related people:

steipete: Current main history shows recent provider mutation and provider catalog work, and this PR branch includes Crownest hardening commits by the same contributor. (role: recent provider and claim-path contributor; confidence: medium; commits: a28b457ccdeb, 185162630e43, 2e5d6dfd559f; files: internal/cli/config.go, internal/providers/all/all.go, internal/providers/crownest/backend.go)
Momo: Added the recent Tencent Cloud provider across the same provider registration, config, metadata, docs, and backend surfaces on current main. (role: adjacent provider contributor; confidence: medium; commits: f18f0ce35887; files: internal/cli/config.go, internal/providers/all/all.go, docs/providers/provider-metadata.json)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3f2a8bfbb3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-06-25T05:33:40Z

+	*shouldStop = false
+	cleanupCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), cleanupTimeout)
+	defer cancel()
+	if err := api.DeleteSandbox(cleanupCtx, sandboxID); err != nil && !isNotFound(err) {


Avoid deleting one-shot Workspace Run sandboxes

For normal crabbox run --provider crownest without --keep, the Workspace Run is created with keepSandbox=false, so Crownest owns cleanup; the public Workspace Runs docs describe the run lifecycle as requiring workspace_run:create/read/cancel and keepSandbox as the cleanup control. This unconditional DeleteSandbox makes an otherwise successful run fail during cleanup for least-privilege keys that can run Workspace Runs but cannot delete sandboxes (403 is not treated like 404), flipping the local exit code to 1 after the remote command succeeded.

Useful? React with 👍 / 👎.

Fixed in e418223. One-shot Crownest Workspace Runs now respect keepSandbox=false: Crabbox removes the local claim but does not call DeleteSandbox, so least-privilege Workspace Run keys do not fail after a successful remote command. Retained paths still delete when Crabbox explicitly requested retention and then needs to clean up.

Validation: go test ./internal/providers/crownest ./internal/cli, go vet ./..., go build -trimpath -o bin/crabbox ./cmd/crabbox, live scripts/live-crownest-smoke.sh, and autoreview clean.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 185162630e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-02T17:18:48Z

+func (c *httpClient) BaseURL() string { return c.baseURL }
+
+func (c *httpClient) Probe(ctx context.Context) error {
+	return c.doJSON(ctx, http.MethodGet, "/v1/sandboxes?limit=1", nil, nil, "")


Probe the Workspace Run surface in doctor

For a least-privilege Crownest key that has only the documented Workspace Run lifecycle scopes (workspace_run:create/read/cancel), normal one-shot crabbox run --provider crownest can succeed, but this doctor probe hits the sandbox-list endpoint and fails with 403 before exercising the delegated-run surface. That makes crabbox doctor --provider crownest incorrectly report auth/API failure for valid run-only credentials; probe a Workspace Run read endpoint instead, or make sandbox-list denial non-fatal for doctor.

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ed1e87e9be

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-02T17:55:49Z

+		if err := b.claimAcquiredSandbox(ctx, api, req, leaseID, sandboxID, slug, &leaseID, &slug); err != nil {
+			return RunResult{}, err
+		}
+		if err := lockAcquiredLease(); err != nil {


Acquire the operation lock before publishing new claims

For newly created Crownest runs, claimAcquiredSandbox writes the local claim and prints the slug before this process holds the per-lease operation lock. In the window before lockAcquiredLease succeeds, another crabbox stop --provider crownest <slug> process can resolve the just-published claim, acquire the lock first, delete the sandbox, and remove the claim while this run is still uploading/finalizing/starting, causing the active run to fail or race against a deleted sandbox. Hold the operation lock before making the claim visible (the same pattern matters when the sandbox ID is first reported after start).

Useful? React with 👍 / 👎.

tristanmanchester added 14 commits June 25, 2026 05:23

feat: add crownest provider

60f0673

fix: harden crownest provider lifecycle

7f921a0

fix: authenticate crownest archive uploads

26fa0df

test: add crownest live smoke

1107b7d

chore: refresh worker lockfile

6bb9a08

fix: harden crownest sync and claim cleanup

0cb2f16

fix: fail crownest canceled runs

f72ac5a

fix: retain crownest setup failures

f74e2a9

fix: clean up crownest claim failures

f95fcd0

fix: redact crownest upload errors

8e4d47d

fix: scope crownest cleanup command

d95703c

chore: regenerate provider categories

ef68512

fix: cancel crownest runs on stream failure

c07c99d

fix: honor crownest status and cleanup state

b824e6f

tristanmanchester marked this pull request as ready for review June 25, 2026 05:28

chatgpt-codex-connector Bot reviewed Jun 25, 2026

View reviewed changes

fix: trim crownest provider surface

e418223

tristanmanchester force-pushed the codex/crownest-provider branch from 3f2a8bf to e418223 Compare June 25, 2026 13:48

clawsweeper Bot added the merge-risk: 🚨 other 🚨 Merging this PR has meaningful risk outside the owned taxonomy. label Jun 25, 2026

steipete added 3 commits July 2, 2026 17:58

chore: merge current main into Crownest PR

84005f7

fix(crownest): clean partial workspace runs

2e5d6df

fix(crownest): harden archive transfer scope

1851626

chatgpt-codex-connector Bot reviewed Jul 2, 2026

View reviewed changes

steipete added 2 commits July 2, 2026 18:37

chore: merge latest main into CROWN candidate

3b123b8

test(crownest): cover configuration paths

ed1e87e

chatgpt-codex-connector Bot reviewed Jul 2, 2026

View reviewed changes

steipete merged commit 30f09ed into openclaw:main Jul 2, 2026
8 checks passed

tristanmanchester deleted the codex/crownest-provider branch July 2, 2026 18:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[codex] Add Crownest provider#681

[codex] Add Crownest provider#681
steipete merged 20 commits into
openclaw:mainfrom
tristanmanchester:codex/crownest-provider

tristanmanchester commented Jun 25, 2026 •

edited by steipete

Loading

Uh oh!

clawsweeper Bot commented Jun 25, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jun 25, 2026

Uh oh!

tristanmanchester Jun 25, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jul 2, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jul 2, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

tristanmanchester commented Jun 25, 2026 • edited by steipete Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Why / impact

Review comment addressed

Repo guidance checked

Validation

Live Crownest evidence

Config / secrets

Uh oh!

clawsweeper Bot commented Jun 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jun 25, 2026

Choose a reason for hiding this comment

Uh oh!

tristanmanchester Jun 25, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

tristanmanchester commented Jun 25, 2026 •

edited by steipete

Loading

clawsweeper Bot commented Jun 25, 2026 •

edited

Loading