fix(connect): preserve DCode observability during probes

[apurvvkumaria]

Summary

Make DCode inference-route health probes preserve managed observability by running them through a side-effect-free, image-owned managed-exec boundary. The new private launcher retains the trusted proxy and shell hardening from #6497, while updated clients fail closed against older images where the helper is absent.

Related Issue

Fixes #6504

Follow-up to #6497, #6412, and #6192.

Changes

• Install a root-owned, non-symlink dcode-managed-exec copy of the reviewed DCode launcher and verify its ownership, mode, contents, and real-path execution during image build.
• Route shared DCode status, doctor, rebuild-preflight, and connect health probes through that side-effect-free boundary instead of the stateful sandbox entrypoint.
• Preserve enabled and disabled observability marker state while retaining managed proxy normalization, startup-file isolation, curl -q, strict output parsing, and old-image fail-closed behavior.
• Add focused launcher, image-contract, shared health, and connect-flow regression coverage.

Type of Change

• Code change (feature, bug fix, or refactor)
• Code change with doc updates
• Doc only (prose changes, no code sample modifications)
• Doc only (includes code sample changes)

Quality Gates

• Tests added or updated for changed behavior
• Existing tests cover changed behavior — justification:
• Tests not applicable — justification:
• Docs updated for user-facing behavior changes
• Docs not applicable — justification: this restores configured DCode observability across internal read-only route probes without changing CLI syntax, output, configuration, or documented behavior.
• Sensitive paths changed (security, policy, credentials, preflight, onboarding, inference, runner, sandbox, or messaging)
• Sensitive-path review completed or maintainer-approved waiver recorded — reviewer/approval link/justification: nine-category security review passed on the exact eight-file diff with no findings; exact-head typed DCode and inference-routing live validation remain required before merge.
• Non-success, skipped, or missing CI check accepted by maintainer — check name, approval link, and follow-up issue:

Verification

• PR description includes the DCO sign-off declaration and every commit appears as Verified in GitHub
• Normal pre-commit, commit-msg, and pre-push hooks passed, or npm run check:diff passed when hooks were skipped or unavailable
• Targeted behavior tests pass for the current change set, or tests are marked not applicable above — command/result or justification: npx vitest run --project cli over the four route/connect/health files passed 61 tests; npx vitest run --project integration over the managed-exec/proxy/image files passed 33 tests.
• Applicable broad gate passed — npm test for broad runtime/test-harness changes; npm run check for repo-wide validation/coverage changes — command/result: not applicable; this narrow boundary fix passed CLI build/typecheck, Bash syntax, test-size, source-shape, title, project-membership, and scoped prek gates.
• Quality Gates section completed with required justifications or waivers
• No secrets, API keys, or credentials committed
• npm run docs builds without warnings (doc changes only)
• Doc pages follow the style guide (doc changes only)
• New doc pages include SPDX header and frontmatter (new doc pages only)

---

Signed-off-by: Apurv Kumaria akumaria@nvidia.com

Summary by CodeRabbit

• New Features

  • Added a managed execution entrypoint for Deep Agents Code so route checks and runtime probing use a dedicated, image-baked launcher path.

• Bug Fixes

  • Improved the launcher’s behavior to avoid altering sandbox state during route diagnostics and to fail safely when mis-invoked.
  • Updated sandbox probe and health-check commands to use the managed entrypoint consistently.

• Tests

  • Added dedicated coverage for the managed entrypoint’s side effects and failure handling.
  • Extended image and sandbox contract tests to validate the managed binary installation and command structure.
  • Improved Deep Agents Code Tavily opt-in checks to reliably restore observability state.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 39c10731-b75a-4fd3-a3ba-06bbc9364de7

📥 Commits

Reviewing files that changed from the base of the PR and between a5dcd02 and fa54bb2.

📒 Files selected for processing (3)

• test/dcode-managed-exec.test.ts
• test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh
• test/e2e/support/platform-parity-cloud-experimental.test.ts

---

📝 Walkthrough

Walkthrough

This PR adds a dcode-managed-exec entry point to the Deep Agents Code launcher, verifies it during image build, repoints sandbox route probes to use it, and adds E2E cleanup so Tavily opt-in checks restore managed observability state.

Changes

Managed exec launcher and build validation

Layer / File(s)	Summary
Launcher managed-exec dispatch agents/langchain-deepagents-code/dcode-launcher.sh, agents/langchain-deepagents-code/Dockerfile, test/langchain-deepagents-code-image.test.ts, test/dcode-managed-exec.test.ts	Adds MANAGED_EXEC_LAUNCHER dispatch, installs and verifies dcode-managed-exec as a non-symlink root:root 0755 copy of the launcher, runs a build-time probe, and adds launcher tests for command handling and observability-marker preservation.
Launcher fixture coverage test/dcode-managed-exec.test.ts	Adds fixture setup and cases for observability-enabled, observability-disabled, and no-command execution paths.

Route probe path updates

Layer / File(s)	Summary
Inference route probe repointed src/lib/actions/sandbox/connect-inference-route-probe.ts, src/lib/actions/sandbox/connect-flow.test.ts, src/lib/actions/sandbox/connect-inference-route-probe.test.ts, src/lib/actions/sandbox/inference-route-health.test.ts	Switches the managed runtime launcher path to /usr/local/lib/nemoclaw/dcode-managed-exec and updates probe assertions to expect the new executable and exact openshell command shape.

Tavily opt-in restoration

Layer / File(s)	Summary
Tavily restoration flow test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh, test/e2e/support/platform-parity-cloud-experimental.test.ts	Adds observability-marker inspection, restoration helpers, an EXIT trap, a self-test branch, and platform-parity coverage for restoring managed observability after Tavily policy removal.

Estimated code review effort: 4 (Complex) | ~45 minutes

Suggested labels: area: sandbox, area: e2e

Suggested reviewers: jyaunches, cv

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: preserving DCode observability during probes.
Linked Issues check	✅ Passed	The changes align with #6504 by routing probes through a side-effect-free managed exec boundary and preserving observability state.
Out of Scope Changes check	✅ Passed	The added Docker, launcher, and test updates all support the probe-observability fix and do not show unrelated scope creep.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/6504-preserve-dcode-observability

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-code-quality]

Code Coverage Overview

Languages: TypeScript

TypeScript / code-coverage/plugin

The overall coverage remains at 96%, unchanged from the main branch.

TypeScript / code-coverage/cli

The overall coverage in the codex/6504-preserve-... branch remains at 76%, unchanged from the main branch.

Show a code coverage summary of the most impacted files.

File	main 10091a3	codex/6504-preserve-... fa54bb2	+/-
src/lib/host-ar...fact-cleanup.ts	83%	17%	-66%
src/lib/actions...-add-restart.ts	19%	14%	-5%
src/lib/agent/onboard.ts	71%	70%	-1%
src/lib/actions...ridge-policy.ts	62%	64%	+2%
src/lib/actions...lution-probe.ts	88%	94%	+6%
src/lib/agent/t...ersion-drift.ts	88%	96%	+8%
src/lib/actions...e-validation.ts	81%	90%	+9%
src/lib/actions...x/mcp-bridge.ts	35%	44%	+9%
src/lib/agent/b...availability.ts	63%	74%	+11%
src/lib/sandbox...rsion-scheme.ts	73%	100%	+27%

---

_{Updated July 08, 2026 23:22 UTC
Code Coverage is in Public Preview. Learn more and provide us with your feedback.}

[github-actions]

E2E Advisor Recommendation

Required E2E: ubuntu-repo-cloud-langchain-deepagents-code, inference-routing
Optional E2E: cloud-onboard

Dispatch hint: targets=ubuntu-repo-cloud-langchain-deepagents-code,inference-routing

Workflow run

Full advisor summary

E2E Recommendation Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required E2E

• ubuntu-repo-cloud-langchain-deepagents-code (high): Required because the PR changes the Deep Agents Code sandbox image, launcher, managed exec boundary, network policy check behavior, observability state handling, and real DCode assistant flows. This existing live target onboards the LangChain Deep Agents Code sandbox from the repo and runs the Deep Agents cloud-experimental checks, including headless inference, Tavily opt-in, TUI startup, observability, and related security-policy checks.
• inference-routing (medium): Required because the PR changes the sandbox inference route probe path and route health/connect diagnostics. The live inference-routing job provides end-to-end confidence that provider routing and inference.local health classification still work with real hosted inference.

Optional E2E

• cloud-onboard (high): Optional broad confidence check for the hosted onboarding/install path. The more precise required Deep Agents Code target should be merge-blocking for this PR, but cloud-onboard can add coverage for general hosted onboarding regressions if maintainers want extra assurance.

New E2E recommendations

• DCode route diagnostics state preservation (medium): Existing coverage is split between unit tests, Dockerfile build checks, and the broad Deep Agents Code live target. A focused live regression would directly prove that connect/status/doctor route probes invoke /usr/local/lib/nemoclaw/dcode-managed-exec through OpenShell exec, preserve the managed observability marker, and do not run the stateful sandbox entrypoint.
  • Suggested test: Add a focused live Deep Agents Code route-probe test or cloud-experimental check that toggles observability, runs the DCode inference route probe via the CLI path, and asserts the marker and proxy-normalized environment remain unchanged.

Dispatch hint

• Workflow: .github/workflows/e2e.yaml
• jobs input: targets=ubuntu-repo-cloud-langchain-deepagents-code,inference-routing

[github-actions]

E2E Target Recommendation

Required E2E targets: ubuntu-repo-cloud-langchain-deepagents-code
Optional E2E targets: None

Dispatch required E2E targets:

• gh workflow run e2e.yaml --ref <pr-head-ref> --field targets=ubuntu-repo-cloud-langchain-deepagents-code

Workflow run

Full E2E target advisor summary

E2E Target Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required E2E targets

• ubuntu-repo-cloud-langchain-deepagents-code: PR changes the LangChain Deep Agents Code image/launcher, the DCode-specific managed inference route probe path, and the cloud-experimental Tavily opt-in E2E check. The smallest live-supported typed target that builds/onboards the DCode sandbox and runs the affected DCode policy/terminal/inference surfaces is ubuntu-repo-cloud-langchain-deepagents-code.
  • Dispatch: gh workflow run e2e.yaml --ref <pr-head-ref> --field targets=ubuntu-repo-cloud-langchain-deepagents-code

Optional E2E targets

• None.

Relevant changed files

• agents/langchain-deepagents-code/Dockerfile
• agents/langchain-deepagents-code/dcode-launcher.sh
• src/lib/actions/sandbox/connect-inference-route-probe.ts
• test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh
• test/e2e/support/platform-parity-cloud-experimental.test.ts

[github-actions]

E2E Target Results — ✅ All requested jobs passed

Run: 28981357492
Workflow ref: codex/6504-preserve-dcode-observability
Requested targets: (default — all supported)
Requested jobs: inference-routing
Summary: 1 passed, 0 failed, 0 cancelled, 0 skipped

Job	Result
inference-routing	✅ success

[github-actions]

PR Review Advisor — Changes requested

Merge posture: Do not merge yet
Primary next action: Resolve or justify PRA-1: Prove or constrain direct dcode-managed-exec egress.
Open items: 0 required · 1 warning · 0 suggestions · 8 test follow-ups
Since last review: 0 prior items resolved · 5 still apply · 1 new item found

Action checklist

• PRA-1 Resolve or justify: Prove or constrain direct dcode-managed-exec egress in agents/langchain-deepagents-code/dcode-launcher.sh:123
• PRA-T1 Add or justify test follow-up: Runtime validation
• PRA-T2 Add or justify test follow-up: Runtime validation
• PRA-T3 Add or justify test follow-up: Runtime validation
• PRA-T4 Add or justify test follow-up: Runtime validation
• PRA-T5 Add or justify test follow-up: Runtime validation
• PRA-T6 Add or justify test follow-up: Acceptance clause
• PRA-T7 Add or justify test follow-up: Acceptance clause
• PRA-T8 Add or justify test follow-up: Acceptance clause

Findings index

ID	Severity	Category	Location	Required action
PRA-1	Resolve/justify	security	agents/langchain-deepagents-code/dcode-launcher.sh:123	Either restrict managed-exec to the expected diagnostic command shape, or add targeted policy evidence showing direct invocation of `/usr/local/lib/nemoclaw/dcode-managed-exec` with arbitrary network commands still fails closed for unapproved egress while the fixed `inference.local` route probe continues to work.

Review findings by urgency: 0 required fixes, 1 item to resolve/justify, 0 in-scope improvements

⚠️ Resolve or justify before merge

Investigate these in the current review; either fix them, explain why they are not applicable, or document the accepted risk.

PRA-1 Resolve/justify — Prove or constrain direct dcode-managed-exec egress

• Location: agents/langchain-deepagents-code/dcode-launcher.sh:123
• Category: security
• Problem: The new `/usr/local/lib/nemoclaw/dcode-managed-exec` mode normalizes the image-baked proxy and observability environment, then `exec "$@"`. Because the Dockerfile installs this copy as root-owned but world-executable `0755`, a sandbox user can invoke the helper directly with arbitrary commands, not only the fixed route-health probe. The intended probe path is safe and array-built, but this broader public entrypoint needs either a narrow command contract or regression evidence that direct use cannot bypass NemoClaw network policy, SSRF controls, or Tavily opt-in restrictions.
• Impact: If the managed proxy or OpenShell policy treats this helper path differently from normal user commands, direct invocation could provide a policy-bypass or SSRF primitive for unapproved hosts. Even if lower layers already deny it, the PR currently does not prove that boundary for the newly exposed executable.
• Recommended action: Either restrict managed-exec to the expected diagnostic command shape, or add targeted policy evidence showing direct invocation of `/usr/local/lib/nemoclaw/dcode-managed-exec` with arbitrary network commands still fails closed for unapproved egress while the fixed `inference.local` route probe continues to work.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Read `agents/langchain-deepagents-code/dcode-launcher.sh` around the `$0 == "$MANAGED_EXEC_LAUNCHER"` branch and `agents/langchain-deepagents-code/Dockerfile` around the `install -o root -g root -m 0755 ... dcode-managed-exec` line; confirm the helper is callable by the sandbox user and accepts arbitrary argv.
• Missing regression test: Add a live or policy-level regression that runs `/usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/python3 -c '...'` or `/usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/curl ...` against an unapproved host such as Tavily before opt-in or another denied endpoint, and asserts the request is blocked; keep a positive assertion that the fixed `https://inference.local/v1/models\` route probe still succeeds or returns the trusted OK/BROKEN grammar.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Read `agents/langchain-deepagents-code/dcode-launcher.sh` around the `$0 == "$MANAGED_EXEC_LAUNCHER"` branch and `agents/langchain-deepagents-code/Dockerfile` around the `install -o root -g root -m 0755 ... dcode-managed-exec` line; confirm the helper is callable by the sandbox user and accepts arbitrary argv.
• Evidence: `dcode-launcher.sh` now branches on `$0` and executes arbitrary `"$@"` after proxy normalization; `Dockerfile` installs `/usr/local/lib/nemoclaw/dcode-managed-exec` with mode `0755`; added tests cover marker preservation and no-command failure but not arbitrary direct egress denial.

💡 In-scope improvements

These are lower-risk, not throwaway. Prefer fixing them in this PR when they are local to changed code; defer only with rationale or a linked follow-up.

• None.

Test follow-ups to resolve or justify

If these cover changed behavior, prefer adding them in this PR; otherwise state why existing coverage is enough or link the follow-up.

• PRA-T1 Runtime validation — In an exact-head DCode sandbox with managed observability enabled, run status, doctor, and connect route-health probes and verify `/tmp/nemoclaw-observability-enabled` remains present before and after each probe.. Static and unit coverage is strong for argv construction, parser behavior, launcher marker preservation, and local E2E cleanup. The changed Dockerfile/image/runtime path, OpenShell exec environment, network policy enforcement, and observability marker state still require behavioral validation in an exact-head sandbox.
• PRA-T2 Runtime validation — In an exact-head DCode sandbox with managed observability disabled, run status, doctor, and connect route-health probes and verify `/tmp/nemoclaw-observability-enabled` is not created.. Static and unit coverage is strong for argv construction, parser behavior, launcher marker preservation, and local E2E cleanup. The changed Dockerfile/image/runtime path, OpenShell exec environment, network policy enforcement, and observability marker state still require behavioral validation in an exact-head sandbox.
• PRA-T3 Runtime validation — Run an exact-head DCode route probe with `inference.local` unreachable and confirm it fails closed through `/usr/local/lib/nemoclaw/dcode-managed-exec` without executing user startup files, trusting inherited FD spoof output, or reading user `.curlrc`.. Static and unit coverage is strong for argv construction, parser behavior, launcher marker preservation, and local E2E cleanup. The changed Dockerfile/image/runtime path, OpenShell exec environment, network policy enforcement, and observability marker state still require behavioral validation in an exact-head sandbox.
• PRA-T4 Runtime validation — Run a non-DCode OpenClaw route-health probe and confirm its argv and behavior remain the plain `sh -c` probe path.. Static and unit coverage is strong for argv construction, parser behavior, launcher marker preservation, and local E2E cleanup. The changed Dockerfile/image/runtime path, OpenShell exec environment, network policy enforcement, and observability marker state still require behavioral validation in an exact-head sandbox.
• PRA-T5 Runtime validation — Directly invoke `/usr/local/lib/nemoclaw/dcode-managed-exec` with an arbitrary command that attempts unapproved egress, such as Tavily before opt-in, and verify OpenShell policy or the managed proxy denies the request.. Static and unit coverage is strong for argv construction, parser behavior, launcher marker preservation, and local E2E cleanup. The changed Dockerfile/image/runtime path, OpenShell exec environment, network policy enforcement, and observability marker state still require behavioral validation in an exact-head sandbox.
• PRA-T6 Acceptance clause — Route-health probes preserve an existing managed observability marker and do not enable observability when it was disabled. — add test evidence or identify existing coverage. `dcode-launcher.sh` unsets ambient `NEMOCLAW_OBSERVABILITY`, re-enables it only from a regular non-symlink marker containing `1`, and managed-exec avoids `start.sh` marker mutation. `test/dcode-managed-exec.test.ts` covers enabled marker preservation, disabled marker non-creation, and no-command preservation. Exact-head live status/doctor/connect behavior remains a runtime-validation follow-up.
• PRA-T7 Acceptance clause — Regression coverage exercises the real managed-launcher boundary and the status/doctor/connect probe call sites. — add test evidence or identify existing coverage. `test/dcode-managed-exec.test.ts` spawns a fixture built from the real `agents/langchain-deepagents-code/dcode-launcher.sh`; `connect-flow.test.ts` covers the connect probe argv; `inference-route-health.test.ts` covers the shared status/doctor health helper argv. There is no separate top-level status/doctor test asserting the exact argv, and live runtime validation remains recommended.
• PRA-T8 Acceptance clause — The typed DCode live target and `inference-routing` live validation pass on the corrective PR head. — add test evidence or identify existing coverage. This clause requires live external validation. Static review did not evaluate external E2E job state; the relevant exact-head runtime checks remain listed under test-depth follow-ups.

Since last review details

Current findings, using the urgency labels above:

PRA-1 Resolve/justify — Prove or constrain direct dcode-managed-exec egress

• Location: agents/langchain-deepagents-code/dcode-launcher.sh:123
• Category: security
• Problem: The new `/usr/local/lib/nemoclaw/dcode-managed-exec` mode normalizes the image-baked proxy and observability environment, then `exec "$@"`. Because the Dockerfile installs this copy as root-owned but world-executable `0755`, a sandbox user can invoke the helper directly with arbitrary commands, not only the fixed route-health probe. The intended probe path is safe and array-built, but this broader public entrypoint needs either a narrow command contract or regression evidence that direct use cannot bypass NemoClaw network policy, SSRF controls, or Tavily opt-in restrictions.
• Impact: If the managed proxy or OpenShell policy treats this helper path differently from normal user commands, direct invocation could provide a policy-bypass or SSRF primitive for unapproved hosts. Even if lower layers already deny it, the PR currently does not prove that boundary for the newly exposed executable.
• Recommended action: Either restrict managed-exec to the expected diagnostic command shape, or add targeted policy evidence showing direct invocation of `/usr/local/lib/nemoclaw/dcode-managed-exec` with arbitrary network commands still fails closed for unapproved egress while the fixed `inference.local` route probe continues to work.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Read `agents/langchain-deepagents-code/dcode-launcher.sh` around the `$0 == "$MANAGED_EXEC_LAUNCHER"` branch and `agents/langchain-deepagents-code/Dockerfile` around the `install -o root -g root -m 0755 ... dcode-managed-exec` line; confirm the helper is callable by the sandbox user and accepts arbitrary argv.
• Missing regression test: Add a live or policy-level regression that runs `/usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/python3 -c '...'` or `/usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/curl ...` against an unapproved host such as Tavily before opt-in or another denied endpoint, and asserts the request is blocked; keep a positive assertion that the fixed `https://inference.local/v1/models\` route probe still succeeds or returns the trusted OK/BROKEN grammar.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Read `agents/langchain-deepagents-code/dcode-launcher.sh` around the `$0 == "$MANAGED_EXEC_LAUNCHER"` branch and `agents/langchain-deepagents-code/Dockerfile` around the `install -o root -g root -m 0755 ... dcode-managed-exec` line; confirm the helper is callable by the sandbox user and accepts arbitrary argv.
• Evidence: `dcode-launcher.sh` now branches on `$0` and executes arbitrary `"$@"` after proxy normalization; `Dockerfile` installs `/usr/local/lib/nemoclaw/dcode-managed-exec` with mode `0755`; added tests cover marker preservation and no-command failure but not arbitrary direct egress denial.

Workflow run details

This is an automated, non-binding review; it still expects maintainers and agents to respond to each required or warning item. Treat suggestions as current-PR improvements when they touch changed code; defer only with maintainer rationale or a linked follow-up. A human maintainer must make the final merge decision.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@test/dcode-managed-exec.test.ts`:
- Around line 68-108: The parameterized test in dcode-managed-exec.test.ts now
adds two conditional branches that trigger the guardrail failure. Split the
current it.each case into two explicit tests for the enabled and disabled
observability scenarios, and move the marker setup/assertion logic into each
test so there are no if statements. Keep the same coverage by preserving the
existing launcher fixture setup, spawnSync invocation, and checks around
makeLauncherFixture, markerPath, wrapperMarkerPath, and the observability env
behavior.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: ebef184e-dd79-46bc-8e11-9ee735c846c2

📥 Commits

Reviewing files that changed from the base of the PR and between 10091a3 and a5dcd02.

📒 Files selected for processing (8)

• agents/langchain-deepagents-code/Dockerfile
• agents/langchain-deepagents-code/dcode-launcher.sh
• src/lib/actions/sandbox/connect-flow.test.ts
• src/lib/actions/sandbox/connect-inference-route-probe.test.ts
• src/lib/actions/sandbox/connect-inference-route-probe.ts
• src/lib/actions/sandbox/inference-route-health.test.ts
• test/dcode-managed-exec.test.ts
• test/langchain-deepagents-code-image.test.ts

[github-actions]

PR Review Advisor (Nemotron Ultra) — Informational

Merge posture: Informational / low confidence
Primary next action: Fix PRA-2: Review finding; then add or justify PRA-T1.
Open items: 2 required · 4 warnings · 2 suggestions · 1 test follow-up
Since last review: 0 prior items resolved · 1 still applies · 3 new items found

Action checklist

• PRA-2 Fix: Review finding in agents/langchain-deepagents-code/dcode-launcher.sh:126-133
• PRA-3 Fix: Review finding in agents/langchain-deepagents-code/Dockerfile:53-57,117
• PRA-1 Resolve or justify: PR review advisor retry failed
• PRA-4 Resolve or justify: Review finding in src/lib/actions/sandbox/connect-inference-route-probe.ts:51
• PRA-5 Resolve or justify: Review finding in test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh:61-102
• PRA-6 Resolve or justify: Review finding in agents/langchain-deepagents-code/Dockerfile:1
• PRA-T1 Add or justify test follow-up: Review finding
• PRA-7 In-scope improvement: Review finding in test/dcode-managed-exec.test.ts
• PRA-8 In-scope improvement: Review finding in src/lib/actions/sandbox/connect-inference-route-probe.ts:51

Findings index

ID	Severity	Category	Location	Required action
PRA-1	Resolve/justify	workflow	—	Treat this result as lower confidence, inspect the raw retry artifact, and rerun the advisor if the preserved findings are unclear.
PRA-2	Required	security	agents/langchain-deepagents-code/dcode-launcher.sh:126-133	No action needed — security posture improved. Verify wrapperMarkerPath is not created in test/dcode-managed-exec.test.ts lines 63, 91, 121.
PRA-3	Required	security	agents/langchain-deepagents-code/Dockerfile:53-57,117	No action needed — version-skew protection correctly implemented. Consider adding build-time probe smoke test with mock curl (PRA-2 from prior review) for deeper integration confidence.
PRA-4	Resolve/justify	security	src/lib/actions/sandbox/connect-inference-route-probe.ts:51	Rename constant to DCODE_MANAGED_EXEC_LAUNCHER in connect-inference-route-probe.ts and update comment. Low risk, local to changed code.
PRA-5	Resolve/justify	security	test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh:61-102	No action needed — proper cleanup and state restoration. Verify self-test mode 'restore-denial' exercises the restoration path (lines 104-123).
PRA-6	Resolve/justify	workflow	agents/langchain-deepagents-code/Dockerfile:1	Coordinate with authors of overlapping PRs. Rebase or resolve conflicts before merge. This is a process coordination item, not a code defect.
PRA-7	Improvement	tests	test/dcode-managed-exec.test.ts	Add integration test in test/e2e/ or test/integration/ that builds the image and invokes the managed-exec entry point directly. Prefer current-PR fix when local to changed code; defer only with rationale.
PRA-8	Improvement	architecture	src/lib/actions/sandbox/connect-inference-route-probe.ts:51	Rename constant for clarity. Safe simplification — only constant rename, all call sites updated.

🚨 Required before merge

Address these before merging unless a maintainer explicitly overrides the advisor with rationale.

PRA-2 Required — Review finding

• Location: agents/langchain-deepagents-code/dcode-launcher.sh:126-133
• Category: security
• Problem: Managed exec mode correctly implements side-effect-free probe path: validates proxy from root-owned files, preserves observability marker (does not consume it), executes command directly without wrapper invocation, and fails closed (exit 64) when no command provided.
• Impact: No impact provided.
• Required action: No action needed — security posture improved. Verify wrapperMarkerPath is not created in test/dcode-managed-exec.test.ts lines 63, 91, 121.
• Expected follow-up: Fix before merge or get explicit maintainer override.
• Verification: Check dcode-launcher.sh lines 126-133: if [ "$0" = "$MANAGED_EXEC_LAUNCHER" ]; then ... exec "$@"; fi — wrapper not invoked. Check test/dcode-managed-exec.test.ts expects wrapperMarkerPath to not exist.
• Missing regression test: test/dcode-managed-exec.test.ts already covers: observability preservation (enabled/disabled), wrapper non-invocation, fail-closed on missing command.
• Done when: The required change is committed and verification passes: Check dcode-launcher.sh lines 126-133: if [ "$0" = "$MANAGED_EXEC_LAUNCHER" ]; then ... exec "$@"; fi — wrapper not invoked. Check test/dcode-managed-exec.test.ts expects wrapperMarkerPath to not exist.
• Evidence: No evidence provided.

PRA-3 Required — Review finding

• Location: agents/langchain-deepagents-code/Dockerfile:53-57,117
• Category: security
• Problem: dcode-managed-exec installed as separate root-owned regular file (not symlink) with validated ownership/mode/content equality to launcher. Build-time test runs `/usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true` proving the entry point works. Older images lack this file → newer CLI probing them fails fast (ENOENT) before stateful entrypoint/wrapper runs.
• Impact: No impact provided.
• Required action: No action needed — version-skew protection correctly implemented. Consider adding build-time probe smoke test with mock curl (PRA-2 from prior review) for deeper integration confidence.
• Expected follow-up: Fix before merge or get explicit maintainer override.
• Verification: Dockerfile lines 53-57: install + test -f + test ! -L + stat 0:0:755 + cmp -s. Line 117: /usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true.
• Missing regression test: PRA-2 suggests build-time probe smoke test with mock curl exercising full probe argv. PRA-3 suggests negative test for version skew: probe argv against missing managed-exec → fail-closed.
• Done when: The required change is committed and verification passes: Dockerfile lines 53-57: install + test -f + test ! -L + stat 0:0:755 + cmp -s. Line 117: /usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true.
• Evidence: No evidence provided.

Review findings by urgency: 2 required fixes, 4 items to resolve/justify, 2 in-scope improvements

⚠️ Resolve or justify before merge

Investigate these in the current review; either fix them, explain why they are not applicable, or document the accepted risk.

PRA-1 Resolve/justify — PR review advisor retry failed

• Location: not file-specific
• Category: workflow
• Problem: The first advisor response parsed, but a quality-improvement retry failed; this result preserves the first-pass review.
• Impact: Maintainers still have the first-pass findings, but low-quality structured fields may remain until a future advisor run succeeds.
• Recommended action: Treat this result as lower confidence, inspect the raw retry artifact, and rerun the advisor if the preserved findings are unclear.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Open pr-review-advisor-retry-raw-output.txt and the workflow logs to inspect the retry failure.
• Missing regression test: Keep unit coverage that proves a retry failure preserves the first normalized review with this limitation.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Open pr-review-advisor-retry-raw-output.txt and the workflow logs to inspect the retry failure.
• Evidence: Could not parse JSON from PR review advisor output; see /home/runner/work/NemoClaw/NemoClaw/artifacts/pr-review-advisor-nemotron-ultra/pr-review-advisor-retry-raw-output.txt

PRA-4 Resolve/justify — Review finding

• Location: src/lib/actions/sandbox/connect-inference-route-probe.ts:51
• Category: security
• Problem: Constant named DCODE_MANAGED_RUNTIME_LAUNCHER but points to dcode-managed-exec (not a runtime launcher). Prior review PRA-4 recommended rename to DCODE_MANAGED_EXEC_LAUNCHER for clarity. Not a security issue but improves code clarity and matches actual file name.
• Impact: No impact provided.
• Recommended action: Rename constant to DCODE_MANAGED_EXEC_LAUNCHER in connect-inference-route-probe.ts and update comment. Low risk, local to changed code.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: grep -n DCODE_MANAGED_RUNTIME_LAUNCHER src/lib/actions/sandbox/connect-inference-route-probe.ts — used at line 51 (def) and 65 (use).
• Missing regression test: None — cosmetic only; existing tests cover behavior.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: grep -n DCODE_MANAGED_RUNTIME_LAUNCHER src/lib/actions/sandbox/connect-inference-route-probe.ts — used at line 51 (def) and 65 (use).
• Evidence: No evidence provided.

PRA-5 Resolve/justify — Review finding

• Location: test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh:61-102
• Category: security
• Problem: Tavily opt-in check adds observability state restoration after policy-remove. Captures marker before policy-remove, restores via `NEMOCLAW_OBSERVABILITY=1 /usr/local/bin/nemoclaw-start /usr/bin/true` if marker was lost. Self-test modes validate restoration logic. Uses trap EXIT for cleanup.
• Impact: No impact provided.
• Recommended action: No action needed — proper cleanup and state restoration. Verify self-test mode 'restore-denial' exercises the restoration path (lines 104-123).
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Check restore_observability_state() function lines 61-83 and restore_tavily_denial() lines 85-102. Self-test at lines 104-123 mocks openshell/nemoclaw_cli to verify marker restoration.
• Missing regression test: Platform parity test added at test/e2e/support/platform-parity-cloud-experimental.test.ts lines 232-247 with 3 parametrized cases for restore-denial logic.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Check restore_observability_state() function lines 61-83 and restore_tavily_denial() lines 85-102. Self-test at lines 104-123 mocks openshell/nemoclaw_cli to verify marker restoration.
• Evidence: No evidence provided.

PRA-6 Resolve/justify — Review finding

• Location: agents/langchain-deepagents-code/Dockerfile:1
• Category: workflow
• Problem: Merge conflict risk with 5 overlapping PRs (fix(dcode): harden Nemotron Ultra tool requests #6494, fix(dcode): route fetch_url through managed proxy #6495, feat(performance): add progressive disclosure tests and compositional routing acceptance #6341, fix(dcode): preserve state across policy reloads #6503, fix(dcode): require Landlock enforcement #5812) modifying same files (Dockerfile, dcode-launcher.sh, test files). Prior review PRA-5 flagged this. This PR's changes are minimal and localized (new entry point + probe switch), so rebasing should be straightforward.
• Impact: No impact provided.
• Recommended action: Coordinate with authors of overlapping PRs. Rebase or resolve conflicts before merge. This is a process coordination item, not a code defect.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Check git merge-base and diff against each overlapping PR branch: `git fetch origin pull/6341/head:pr6341 && git diff pr6341...HEAD -- agents/langchain-deepagents-code/Dockerfile agents/langchain-deepagents-code/dcode-launcher.sh`
• Missing regression test: N/A — process coordination.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Check git merge-base and diff against each overlapping PR branch: `git fetch origin pull/6341/head:pr6341 && git diff pr6341...HEAD -- agents/langchain-deepagents-code/Dockerfile agents/langchain-deepagents-code/dcode-launcher.sh`.
• Evidence: No evidence provided.

💡 In-scope improvements

These are lower-risk, not throwaway. Prefer fixing them in this PR when they are local to changed code; defer only with rationale or a linked follow-up.

PRA-7 Improvement — Review finding

• Location: test/dcode-managed-exec.test.ts
• Category: tests
• Problem: Unit tests use string replacement on launcher source for isolation. An integration test building the DCode image and running `dcode-managed-exec /usr/bin/true` and `dcode-managed-exec /bin/sh -c 'exit 0'` in-container would verify the actual installed entry point works end-to-end (PRA-1 from prior review).
• Impact: No impact provided.
• Suggested action: Add integration test in test/e2e/ or test/integration/ that builds the image and invokes the managed-exec entry point directly. Prefer current-PR fix when local to changed code; defer only with rationale.
• Expected follow-up: Prefer a current-PR fix when local to changed code; defer only with rationale or linked follow-up.
• Verification: Check if test/e2e/ has a DCode image build + probe test; run `docker build -t nemoclaw-dcode-test agents/langchain-deepagents-code && docker run --rm nemoclaw-dcode-test /usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true`
• Missing regression test: Integration test exercising actual Docker image entry point for dcode-managed-exec
• Done when: The local improvement is applied, or the PR notes why it should be deferred.
• Evidence: No evidence provided.

PRA-8 Improvement — Review finding

• Location: src/lib/actions/sandbox/connect-inference-route-probe.ts:51
• Category: architecture
• Problem: Simplification opportunity: rename constant DCODE_MANAGED_RUNTIME_LAUNCHER to DCODE_MANAGED_EXEC_LAUNCHER (shrink — net 0 lines, no behavioral change). Prior review PRA-4 identified this.
• Impact: No impact provided.
• Suggested action: Rename constant for clarity. Safe simplification — only constant rename, all call sites updated.
• Expected follow-up: Prefer a current-PR fix when local to changed code; defer only with rationale or linked follow-up.
• Verification: grep -n DCODE_MANAGED_RUNTIME_LAUNCHER src/lib/actions/sandbox/connect-inference-route-probe.ts
• Missing regression test: None — cosmetic only; existing tests cover behavior.
• Done when: The local improvement is applied, or the PR notes why it should be deferred.
• Evidence: No evidence provided.

Test follow-ups to resolve or justify

If these cover changed behavior, prefer adding them in this PR; otherwise state why existing coverage is enough or link the follow-up.

• PRA-T1 Review finding — Add integration test in test/e2e/ or test/integration/ that builds the image and invokes the managed-exec entry point directly. Prefer current-PR fix when local to changed code; defer only with rationale.

Since last review details

Current findings, using the urgency labels above:

PRA-1 Resolve/justify — PR review advisor retry failed

• Location: not file-specific
• Category: workflow
• Problem: The first advisor response parsed, but a quality-improvement retry failed; this result preserves the first-pass review.
• Impact: Maintainers still have the first-pass findings, but low-quality structured fields may remain until a future advisor run succeeds.
• Recommended action: Treat this result as lower confidence, inspect the raw retry artifact, and rerun the advisor if the preserved findings are unclear.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Open pr-review-advisor-retry-raw-output.txt and the workflow logs to inspect the retry failure.
• Missing regression test: Keep unit coverage that proves a retry failure preserves the first normalized review with this limitation.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Open pr-review-advisor-retry-raw-output.txt and the workflow logs to inspect the retry failure.
• Evidence: Could not parse JSON from PR review advisor output; see /home/runner/work/NemoClaw/NemoClaw/artifacts/pr-review-advisor-nemotron-ultra/pr-review-advisor-retry-raw-output.txt

PRA-2 Required — Review finding

• Location: agents/langchain-deepagents-code/dcode-launcher.sh:126-133
• Category: security
• Problem: Managed exec mode correctly implements side-effect-free probe path: validates proxy from root-owned files, preserves observability marker (does not consume it), executes command directly without wrapper invocation, and fails closed (exit 64) when no command provided.
• Impact: No impact provided.
• Required action: No action needed — security posture improved. Verify wrapperMarkerPath is not created in test/dcode-managed-exec.test.ts lines 63, 91, 121.
• Expected follow-up: Fix before merge or get explicit maintainer override.
• Verification: Check dcode-launcher.sh lines 126-133: if [ "$0" = "$MANAGED_EXEC_LAUNCHER" ]; then ... exec "$@"; fi — wrapper not invoked. Check test/dcode-managed-exec.test.ts expects wrapperMarkerPath to not exist.
• Missing regression test: test/dcode-managed-exec.test.ts already covers: observability preservation (enabled/disabled), wrapper non-invocation, fail-closed on missing command.
• Done when: The required change is committed and verification passes: Check dcode-launcher.sh lines 126-133: if [ "$0" = "$MANAGED_EXEC_LAUNCHER" ]; then ... exec "$@"; fi — wrapper not invoked. Check test/dcode-managed-exec.test.ts expects wrapperMarkerPath to not exist.
• Evidence: No evidence provided.

PRA-3 Required — Review finding

• Location: agents/langchain-deepagents-code/Dockerfile:53-57,117
• Category: security
• Problem: dcode-managed-exec installed as separate root-owned regular file (not symlink) with validated ownership/mode/content equality to launcher. Build-time test runs `/usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true` proving the entry point works. Older images lack this file → newer CLI probing them fails fast (ENOENT) before stateful entrypoint/wrapper runs.
• Impact: No impact provided.
• Required action: No action needed — version-skew protection correctly implemented. Consider adding build-time probe smoke test with mock curl (PRA-2 from prior review) for deeper integration confidence.
• Expected follow-up: Fix before merge or get explicit maintainer override.
• Verification: Dockerfile lines 53-57: install + test -f + test ! -L + stat 0:0:755 + cmp -s. Line 117: /usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true.
• Missing regression test: PRA-2 suggests build-time probe smoke test with mock curl exercising full probe argv. PRA-3 suggests negative test for version skew: probe argv against missing managed-exec → fail-closed.
• Done when: The required change is committed and verification passes: Dockerfile lines 53-57: install + test -f + test ! -L + stat 0:0:755 + cmp -s. Line 117: /usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true.
• Evidence: No evidence provided.

PRA-4 Resolve/justify — Review finding

• Location: src/lib/actions/sandbox/connect-inference-route-probe.ts:51
• Category: security
• Problem: Constant named DCODE_MANAGED_RUNTIME_LAUNCHER but points to dcode-managed-exec (not a runtime launcher). Prior review PRA-4 recommended rename to DCODE_MANAGED_EXEC_LAUNCHER for clarity. Not a security issue but improves code clarity and matches actual file name.
• Impact: No impact provided.
• Recommended action: Rename constant to DCODE_MANAGED_EXEC_LAUNCHER in connect-inference-route-probe.ts and update comment. Low risk, local to changed code.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: grep -n DCODE_MANAGED_RUNTIME_LAUNCHER src/lib/actions/sandbox/connect-inference-route-probe.ts — used at line 51 (def) and 65 (use).
• Missing regression test: None — cosmetic only; existing tests cover behavior.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: grep -n DCODE_MANAGED_RUNTIME_LAUNCHER src/lib/actions/sandbox/connect-inference-route-probe.ts — used at line 51 (def) and 65 (use).
• Evidence: No evidence provided.

PRA-5 Resolve/justify — Review finding

• Location: test/e2e/e2e-cloud-experimental/checks/09-deepagents-code-tavily-opt-in.sh:61-102
• Category: security
• Problem: Tavily opt-in check adds observability state restoration after policy-remove. Captures marker before policy-remove, restores via `NEMOCLAW_OBSERVABILITY=1 /usr/local/bin/nemoclaw-start /usr/bin/true` if marker was lost. Self-test modes validate restoration logic. Uses trap EXIT for cleanup.
• Impact: No impact provided.
• Recommended action: No action needed — proper cleanup and state restoration. Verify self-test mode 'restore-denial' exercises the restoration path (lines 104-123).
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Check restore_observability_state() function lines 61-83 and restore_tavily_denial() lines 85-102. Self-test at lines 104-123 mocks openshell/nemoclaw_cli to verify marker restoration.
• Missing regression test: Platform parity test added at test/e2e/support/platform-parity-cloud-experimental.test.ts lines 232-247 with 3 parametrized cases for restore-denial logic.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Check restore_observability_state() function lines 61-83 and restore_tavily_denial() lines 85-102. Self-test at lines 104-123 mocks openshell/nemoclaw_cli to verify marker restoration.
• Evidence: No evidence provided.

PRA-6 Resolve/justify — Review finding

• Location: agents/langchain-deepagents-code/Dockerfile:1
• Category: workflow
• Problem: Merge conflict risk with 5 overlapping PRs (fix(dcode): harden Nemotron Ultra tool requests #6494, fix(dcode): route fetch_url through managed proxy #6495, feat(performance): add progressive disclosure tests and compositional routing acceptance #6341, fix(dcode): preserve state across policy reloads #6503, fix(dcode): require Landlock enforcement #5812) modifying same files (Dockerfile, dcode-launcher.sh, test files). Prior review PRA-5 flagged this. This PR's changes are minimal and localized (new entry point + probe switch), so rebasing should be straightforward.
• Impact: No impact provided.
• Recommended action: Coordinate with authors of overlapping PRs. Rebase or resolve conflicts before merge. This is a process coordination item, not a code defect.
• Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
• Verification: Check git merge-base and diff against each overlapping PR branch: `git fetch origin pull/6341/head:pr6341 && git diff pr6341...HEAD -- agents/langchain-deepagents-code/Dockerfile agents/langchain-deepagents-code/dcode-launcher.sh`
• Missing regression test: N/A — process coordination.
• Done when: The risk is fixed or explicitly justified in the PR. Verification: Check git merge-base and diff against each overlapping PR branch: `git fetch origin pull/6341/head:pr6341 && git diff pr6341...HEAD -- agents/langchain-deepagents-code/Dockerfile agents/langchain-deepagents-code/dcode-launcher.sh`.
• Evidence: No evidence provided.

PRA-7 Improvement — Review finding

• Location: test/dcode-managed-exec.test.ts
• Category: tests
• Problem: Unit tests use string replacement on launcher source for isolation. An integration test building the DCode image and running `dcode-managed-exec /usr/bin/true` and `dcode-managed-exec /bin/sh -c 'exit 0'` in-container would verify the actual installed entry point works end-to-end (PRA-1 from prior review).
• Impact: No impact provided.
• Suggested action: Add integration test in test/e2e/ or test/integration/ that builds the image and invokes the managed-exec entry point directly. Prefer current-PR fix when local to changed code; defer only with rationale.
• Expected follow-up: Prefer a current-PR fix when local to changed code; defer only with rationale or linked follow-up.
• Verification: Check if test/e2e/ has a DCode image build + probe test; run `docker build -t nemoclaw-dcode-test agents/langchain-deepagents-code && docker run --rm nemoclaw-dcode-test /usr/local/lib/nemoclaw/dcode-managed-exec /usr/bin/true`
• Missing regression test: Integration test exercising actual Docker image entry point for dcode-managed-exec
• Done when: The local improvement is applied, or the PR notes why it should be deferred.
• Evidence: No evidence provided.

PRA-8 Improvement — Review finding

• Location: src/lib/actions/sandbox/connect-inference-route-probe.ts:51
• Category: architecture
• Problem: Simplification opportunity: rename constant DCODE_MANAGED_RUNTIME_LAUNCHER to DCODE_MANAGED_EXEC_LAUNCHER (shrink — net 0 lines, no behavioral change). Prior review PRA-4 identified this.
• Impact: No impact provided.
• Suggested action: Rename constant for clarity. Safe simplification — only constant rename, all call sites updated.
• Expected follow-up: Prefer a current-PR fix when local to changed code; defer only with rationale or linked follow-up.
• Verification: grep -n DCODE_MANAGED_RUNTIME_LAUNCHER src/lib/actions/sandbox/connect-inference-route-probe.ts
• Missing regression test: None — cosmetic only; existing tests cover behavior.
• Done when: The local improvement is applied, or the PR notes why it should be deferred.
• Evidence: No evidence provided.

Workflow run details

This is an automated, non-binding review; it still expects maintainers and agents to respond to each required or warning item. Treat suggestions as current-PR improvements when they touch changed code; defer only with maintainer rationale or a linked follow-up. A human maintainer must make the final merge decision.

[github-actions]

E2E Target Results — ❌ Some jobs failed

Run: 28981352252
Workflow ref: codex/6504-preserve-dcode-observability
Requested targets: ubuntu-repo-cloud-langchain-deepagents-code
Requested jobs: (default — all default-enabled free-standing jobs; explicit-only jobs openshell-gateway-auth-contract, mcp-bridge-dev, hermes-gpu-startup, sandbox-rlimits-connect, and jetson-nvmap-gpu are skipped unless selected)
Summary: 0 passed, 1 failed, 0 cancelled, 0 skipped

Job	Result
live	❌ failure

Failed jobs: live. Check run artifacts for logs.

[github-actions]

E2E Target Results — ✅ All requested jobs passed

Run: 28981807457
Workflow ref: codex/6504-preserve-dcode-observability
Requested targets: (default — all supported)
Requested jobs: inference-routing
Summary: 1 passed, 0 failed, 0 cancelled, 0 skipped

Job	Result
inference-routing	✅ success

[cjagwani]

Exact-head babysitting follow-up: DCode run 28981352252 passed the managed-observability boundary through check 11, then failed check 12 because check 09 left the Tavily opt-in active. Signed commit fa54bb2 ports only the hermetic check-09 cleanup from draft #6503 (including preservation of the pre-cleanup observability marker); #6503's credential and prerequisite changes remain in that draft. Local verification: 20/20 focused platform/managed-exec tests under the fixture's expected umask, source-shape and test-size ratchets, Bash syntax, ShellCheck, shfmt, and Biome. Fresh exact-head runs: DCode 28981806191 and inference-routing 28981807457 (passed); full CI rerun in progress.

[github-actions]

E2E Target Results — ✅ All requested jobs passed

Run: 28982027565
Workflow ref: codex/6504-preserve-dcode-observability
Requested targets: (default — all supported)
Requested jobs: inference-routing
Summary: 1 passed, 0 failed, 0 cancelled, 0 skipped

Job	Result
inference-routing	✅ success

[github-actions]

E2E Target Results — ✅ All selected jobs passed

Run: 28981806191
Workflow ref: codex/6504-preserve-dcode-observability
Requested targets: ubuntu-repo-cloud-langchain-deepagents-code
Requested jobs: (default — all default-enabled free-standing jobs; explicit-only jobs openshell-gateway-auth-contract, mcp-bridge-dev, hermes-gpu-startup, sandbox-rlimits-connect, and jetson-nvmap-gpu are skipped unless selected)
Summary: 1 passed, 0 failed, 0 cancelled, 0 skipped

Job	Result
live	✅ success

[github-actions]

E2E Target Results — ⚠️ Run cancelled — no signal

Run: 28982022601
Workflow ref: codex/6504-preserve-dcode-observability
Requested targets: ubuntu-repo-cloud-langchain-deepagents-code
Requested jobs: (default — all default-enabled free-standing jobs; explicit-only jobs openshell-gateway-auth-contract, mcp-bridge-dev, hermes-gpu-startup, sandbox-rlimits-connect, and jetson-nvmap-gpu are skipped unless selected)
Summary: 0 passed, 0 failed, 1 cancelled, 0 skipped

Job	Result
live	⚠️ cancelled