chore(harness): repin sandbox and code images to the 0.40.0 release digests

[waynesun09]

Summary

Repin the seven harness/*.yaml image references from the 0.37.0 digests to the 0.40.0 release images:

Image	Digest	Files
ghcr.io/fullsend-ai/fullsend-code:0.40.0	sha256:ea2a31f38ee80e2a9a898a898a289fe432aa882fa5a4046c3236ab8e2627d7e7	code, fix, review
ghcr.io/fullsend-ai/fullsend-sandbox:0.40.0	sha256:f8255971fec8a72a60adb20f501e55098f237e699dc829f0894647c0de2a19c2	prioritize, retro, scribe, triage

Both digests were resolved from GHCR and their org.opencontainers.image.revision label is 2a103497560b9bcca901372c70a6e073b1a73cde, the fullsend v0.40.0 release commit (release run succeeded end to end, including tag-agents).

Why

• The 0.38.0 and 0.39.0 repins were never done (release: automate the fullsend-ai/agents harness image digest repin after each release fullsend#6607 tracks automating this), so fleet agents have been running 0.37.0 images.
• 0.40.0 is the first sandbox image built with CODEX_VERSION; the codex runtime shipped in fullsend v0.40.0 is unusable by fleet agents until these pins move.
• Completes the post-release step for the packages:read rollout in fullsend v0.40.0.

Pins reference the release-tagged digests, never :latest, per the release runbook.

[qodo-code-review]

PR Summary by Qodo

Repin harness images to verified 0.40.0 release digests

⚙️ Configuration changes 🕐 Less than 10 minutes

AI Description

• Pins three code harnesses to the verified fullsend-code 0.40.0 digest.
• Pins four workflow harnesses to the verified fullsend-sandbox 0.40.0 digest.
• Enables fleet agents to use the 0.40.0 Codex runtime.

Diagram

graph TD
  FLEET["Fleet agents"] --> CODECFG["Code harnesses"] --> CODE["Code image 0.40.0"]
  FLEET --> SANDCFG["Workflow harnesses"] --> SANDBOX["Sandbox image 0.40.0"]

Loading

High-Level Assessment

The current approach is optimal: pinning verified release digests provides reproducible deployments while enabling the 0.40.0 runtime. Mutable release or latest tags were considered but would weaken provenance and repeatability.

Files changed (7) +7 / -7

Other (7) +7 / -7

code.yamlRepin code harness to fullsend-code 0.40.0 +1/-1

Repin code harness to fullsend-code 0.40.0

• Replaces the 0.37.0 fullsend-code digest with the verified 0.40.0 release digest for code tasks.

harness/code.yaml

fix.yamlRepin fix harness to fullsend-code 0.40.0 +1/-1

Repin fix harness to fullsend-code 0.40.0

• Replaces the 0.37.0 fullsend-code digest with the verified 0.40.0 release digest for fix tasks.

harness/fix.yaml

prioritize.yamlRepin prioritize harness to fullsend-sandbox 0.40.0 +1/-1

Repin prioritize harness to fullsend-sandbox 0.40.0

• Replaces the 0.37.0 sandbox digest with the verified 0.40.0 release digest for prioritization tasks.

harness/prioritize.yaml

retro.yamlRepin retro harness to fullsend-sandbox 0.40.0 +1/-1

Repin retro harness to fullsend-sandbox 0.40.0

• Replaces the 0.37.0 sandbox digest with the verified 0.40.0 release digest for retrospective tasks.

harness/retro.yaml

review.yamlRepin review harness to fullsend-code 0.40.0 +1/-1

Repin review harness to fullsend-code 0.40.0

• Replaces the 0.37.0 fullsend-code digest with the verified 0.40.0 release digest for review tasks.

harness/review.yaml

scribe.yamlRepin scribe harness to fullsend-sandbox 0.40.0 +1/-1

Repin scribe harness to fullsend-sandbox 0.40.0

• Replaces the 0.37.0 sandbox digest with the verified 0.40.0 release digest for scribe tasks.

harness/scribe.yaml

triage.yamlRepin triage harness to fullsend-sandbox 0.40.0 +1/-1

Repin triage harness to fullsend-sandbox 0.40.0

• Replaces the 0.37.0 sandbox digest with the verified 0.40.0 release digest for triage tasks.

harness/triage.yaml

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 3:38 AM UTC · Completed 3:51 AM UTC

Commit: 495e3cb · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $3.20

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📜 Skill insights (1)

1. Protected harness path modified 📜 Skill insight § Compliance

Description

This PR modifies files under the protected harness/ path, which requires mandatory human review
and must not be auto-approved. The digest-only nature of the change does not exempt protected-path
review.

Code

harness/code.yaml[23]

+image: ghcr.io/fullsend-ai/fullsend-code@sha256:ea2a31f38ee80e2a9a898a898a289fe432aa882fa5a4046c3236ab8e2627d7e7

Relevance

●●● Strong

Recent protected-path enforcement precedents support flagging harness changes; digest-only edits
remain governance-sensitive.

PR-#569
PR-#631

ⓘ Recommendations generated based on similar findings in past PRs

Evidence

The PR changes an image reference in harness/code.yaml; harness/ is explicitly listed as a
protected governance/infrastructure path requiring a finding.

harness/code.yaml[23-23]
Skill: pr-review

---

Context sources

✅ Compliance rules (platform): 57 rules

✅ Skills: 4 invoked

  code-review

  code-implementation

  pr-review

  docs-review

Review mode: 🚀 Fast: This is a localized, mechanical update of seven pinned container digests with no source-logic changes, and the stated release-tagged digests make the intended behavior clear.

Tip of the day

💡 Did you know, you can add REVIEW.md to your repo root and Qodo follows it on every PR

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

[fullsend-ai-review]

Risk Assessment: elevated (3/5)

Details

Elevated score driven by 7 protected harness paths touched and high git churn/contention history on these files, despite the change itself being a minimal 1-line-per-file image digest repin with no security, CI, or dependency impact.

[fullsend-ai-review]

Review

Findings

High

• [protected-path] harness/code.yaml, harness/fix.yaml, harness/prioritize.yaml, harness/retro.yaml, harness/review.yaml, harness/scribe.yaml, harness/triage.yaml — All seven changed files are under harness/, a protected path requiring human approval. This PR does not link to an issue authorizing modifications to governance/infrastructure files. The description explains the rationale (repinning images to the v0.40.0 release digests), but protected-path changes require a linked issue for traceability.

---

Next steps:

• /fs-fix — agent addresses review findings automatically
• /fs-fix <your instruction> — agent fixes with your specific guidance
• Push commits directly — review re-runs automatically on push
• /fs-fix-stop — disable automatic fix runs for this PR

[fullsend-ai-review]

See the review comment for full details.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 11:42 AM UTC · Completed 11:53 AM UTC

Commit: 495e3cb · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $4.56

[fullsend-ai-retro]

PR #1160 was a straightforward infrastructure change: repinning 7 harness YAML files from v0.37.0 to v0.40.0 container image digests (7 single-line swaps). The review agent ran on opus at high effort ($3.20), correctly identified the protected-path governance gate, and submitted CHANGES_REQUESTED. A human reviewer (rh-hemartin) approved, and the author merged ~8 hours later. All CI passed on the final run.

Novel finding

The review bot's risk/elevated label application at 03:51:36 UTC triggered a new Functional Tests workflow run via the pull_request_target labeled event. This new run joined concurrency group functional-1160 with cancel-in-progress: true, cancelling the still-running functional-tests (review) job from the original run (33712022619). The new run (33712899566) then immediately failed: its gate job skipped all tests because the label was not ok-to-test, and the rollup script failed closed because the prior result was already failure. This forced a full re-run of functional tests after human approval, adding unnecessary latency. One proposal filed below.

Corroborating evidence for existing open issues

• #741 (assess semantic significance before assigning HIGH severity): this PR is another instance — 7 purely mechanical single-line digest swaps all received HIGH severity protected-path findings.
• #735 (accept PR body rationale as authorization context): the PR body included detailed justification, a digest-to-commit mapping table, and a clear "Why" section, but the review agent still assigned HIGH because no Fixes #N link existed in the agents repo.
• #718 (recognize external issue references): the PR body referenced release: automate the fullsend-ai/agents harness image digest repin after each release fullsend#6607 as the tracking issue for automating harness repins, but the review agent did not recognize cross-repo references as authorization context.
• #1110 (use COMMENT verdict when all findings are human-only): the only finding was protected-path, which is inherently unresolvable by the fix agent — yet CHANGES_REQUESTED was submitted.
• #497 (scope-based model downgrading for trivial PRs): $3.20 on opus/high for a 7-file, 7-line mechanical digest repin. The scope was classified as trivial but sub-agents still ran on opus.

Proposals filed

• Review bot label application cancels in-progress functional tests via concurrency group race (in fullsend-ai/agents)