Deterministic Docker: fake the dependency/server and assert the exact runtime contract crossing the boundary.
Live Docker: use real provider credentials/model only when user-visible behavior needs the real service.
Prefer both when they prove different risks: deterministic for byte/payload routing, live for actual provider behavior.

Authoring Rules

Test-only helpers live in test/helpers or scripts/e2e/lib/<lane>/, not src/**, unless production imports them.
Package-installed app runs from /app; mount only explicit harness/helper paths read-only.
Fake servers should log boundary requests as JSONL and clients should assert the real dependency payload, not just process success.
Add the package script and scripts/lib/docker-e2e-scenarios.mjs lane in the same change.
If a lane installs a plugin from npm, default the spec via env so published and local override paths are both testable.

Media And Vision

Expected answer must exist only in pixels or provider output being tested.
Use neutral filenames, neutral prompts, and no metadata leaks.
Random bitmap/OCR tokens reuse the repo OCR-safe alphabet 24567ACEF unless the test owns a stronger glyph set.
Make the expected answer unique per run when proving real image understanding.

`chat.send` E2E

Require chat.send to return status: "started" and a string runId.
Wait for completion with agent.wait.
Assert final user-visible text via chat.history when event ordering is not the behavior under test.
Keep originating channel/account metadata only when the bug path needs queued inbound/channel context.

Verification

Run the smallest proof that covers the touched lane:

pnpm exec oxfmt --write <changed files>
node --check <new .mjs files>
bash -n <new .sh files>
node scripts/run-vitest.mjs test/scripts/docker-e2e-plan.test.ts
OPENCLAW_SKIP_DOCKER_BUILD=1 pnpm test:docker:<lane>

For real-provider lanes, run the matching live Docker script after deterministic Docker is green. Finish with $autoreview before commit/PR.

Plus depuis ce dépôt

même dépôt

weather

openclaw/openclaw

Current weather and forecasts with web_fetch, falling back to wttr.in curl for locations, rain, temperature, travel planning.

2026-06-05377.9k

openclaw-secret-scanning-maintainer

openclaw/openclaw

Triage, redact, clean up, and resolve OpenClaw GitHub Secret Scanning alerts in issues or PRs.

2026-06-04377.9k

openclaw-test-heap-leaks

openclaw/openclaw

Investigate OpenClaw pnpm test memory growth, Vitest OOMs, RSS spikes, and heap snapshot deltas.

2026-06-04377.9k

release-openclaw-ci

openclaw/openclaw

Run, watch, debug, and summarize OpenClaw full release CI, release checks, live provider gates, install/update proofs, and release-secret preflights.

2026-06-04377.9k

release-openclaw-maintainer

openclaw/openclaw

Prepare or verify OpenClaw stable/beta releases, changelogs, release notes, publish commands, and artifacts.

2026-06-03377.9k

autoreview

openclaw/openclaw

Auto Review closeout. Codex review is the default when no engine is set and is the recommended reviewer.

2026-06-02377.9k

OpenClaw Docker E2E Authoring

Use this when adding or changing Docker E2E lanes, release-path Docker tests, or live-provider Docker proof.

Lane Choice

Deterministic Docker: fake the dependency/server and assert the exact runtime contract crossing the boundary.

Live Docker: use real provider credentials/model only when user-visible behavior needs the real service.

Prefer both when they prove different risks: deterministic for byte/payload routing, live for actual provider behavior.

Authoring Rules

Test-only helpers live in test/helpers or scripts/e2e/lib/<lane>/, not src/**, unless production imports them.

Package-installed app runs from /app; mount only explicit harness/helper paths read-only.

Fake servers should log boundary requests as JSONL and clients should assert the real dependency payload, not just process success.

Add the package script and scripts/lib/docker-e2e-scenarios.mjs lane in the same change.

If a lane installs a plugin from npm, default the spec via env so published and local override paths are both testable.

Media And Vision

Expected answer must exist only in pixels or provider output being tested.

Use neutral filenames, neutral prompts, and no metadata leaks.

Random bitmap/OCR tokens reuse the repo OCR-safe alphabet 24567ACEF unless the test owns a stronger glyph set.

Make the expected answer unique per run when proving real image understanding.

chat.send E2E

Require chat.send to return status: "started" and a string runId.

Wait for completion with agent.wait.

Assert final user-visible text via chat.history when event ordering is not the behavior under test.

Keep originating channel/account metadata only when the bug path needs queued inbound/channel context.

Verification

Run the smallest proof that covers the touched lane:

pnpm exec oxfmt --write <changed files> node --check <new .mjs files> bash -n <new .sh files> node scripts/run-vitest.mjs test/scripts/docker-e2e-plan.test.ts OPENCLAW_SKIP_DOCKER_BUILD=1 pnpm test:docker:<lane>

For real-provider lanes, run the matching live Docker script after deterministic Docker is green. Finish with $autoreview before commit/PR.

openclaw-docker-e2e-authoring