[Copilot]

GPT-5.6 Terra is retired across Copilot on 16 November 2026. Purelock, Front Page Copy Guard and Refactoring Cadence each pin copilot/gpt-5.6-terra and would stop working then. This PR moves them to cost-oriented model routing (economy) instead of pinning another model.

Frontmatter

• purelock.md, front-page-copy-guard.md: remove model: copilot/gpt-5.6-terra and add routing to the existing Codex engine block:

  engine:
    id: codex
    model-provider: github
    model-routing:
      goal: cost
      mode: economy

• refactoring-cadence.md: replace engine: pi and the Terra pin with the same block, using id: pi.
• copilot-requests: write was already present on all three and is kept. Prompts, steps, sub-agents, triggers and safe outputs are unchanged, and no model allow or block lists are added.

Lock-file changes (all three)

• The agent's AWF config enables routing: "routing":{"enabled":true,"goal":"cost","mode":"economy",…}.
• Adds the "Prepare model-routing conversation" step and pins the AWF router image gh-aw-firewall/router:0.29.6@sha256:fce2f3be….
• Inference uses COPILOT_GITHUB_TOKEN: ${{ github.token }}.
• The Terra pin is removed from the agent and from threat detection (COPILOT_MODEL), so detection uses its default model.
• Downstream jobs get GH_AW_MODEL_ROUTING* env.

Notes for reviewers

• New compile warning on Purelock: sub-agent "test-writer" model "large" may be incompatible with the codex engine's model-routing API. The issue asks to keep sub-agents unchanged, so this PR leaves it for a follow-up.
• make agent-finish stops at golint: it flags three style findings in existing pkg/cli/update_*.go files that this PR doesn't touch. I didn't check whether this also fails on main.
• Security scan: the CodeQL check timed out and was not re-run.
• Out of scope:
  • windows.md, cloclo.md and smoke-copilot-sub-agents.md, which the issue lists separately.
  • Purelock's sub-agent startup failures ([aw] PureLock: reported incomplete result #74513), which are unrelated.
• Follow-up: audit the first routed runs with gh aw audit (the issue's "Check" section). The issue expects GPT-5.6 Luna picks, or Claude Haiku 5.5 for pi. Repeated gpt-5.6-sol picks would mean switching to a copilot/gpt-6.1-sol pin.

• Fixes Model routing: move Purelock, Front Page Copy Guard and Refactoring Cadence off GPT-5.6 Terra before it retires on 16 November #74642

[github-actions]

🧠 Matt Pocock Skills Reviewer has completed the skills-based review. ✅

🧠 Reviewed using Matt Pocock's skills by Matt Pocock Skills Reviewer

[github-actions]

✅ PR Code Quality Reviewer completed the code quality review.

🔎 Code quality review by PR Code Quality Reviewer

[github-actions]

✅ Design Decision Gate 🏗️ completed the design decision gate check. See the comment below for the result and any generated ADR draft.

Design Decision Gate 🏗️

Outcome: reported_no_action

The agent completed and reported no action. A no-op output alone does not verify that all required work was completed or that publication was possible.

Agent messages are omitted from this comment. The category uses structured run status, not interpretation of agent prose.

Generated from Design Decision Gate 🏗️ · routed: opus55 medium · ~12 AIC · ⌖ ~2 AIC · ⊞ 7.7K · ◷

🏗️ ADR gate enforced by Design Decision Gate 🏗️

[github-actions]

✅ Test Quality Sentinel completed test quality analysis.

Test Quality Sentinel

Outcome: reported_no_action

The agent completed and reported no action. A no-op output alone does not verify that all required work was completed or that publication was possible.

Agent messages are omitted from this comment. The category uses structured run status, not interpretation of agent prose.

Generated from Test Quality Sentinel · routed: gpt56 max · ~2 AIC · ⌖ ~1 AIC · ⊞ 6.3K · ◷

🧪 Test quality analysis by Test Quality Sentinel

[github-actions]

✅ Ponytail Reviewer completed successfully!

Ponytail Reviewer

Outcome: reported_no_action

The agent completed and reported no action. A no-op output alone does not verify that all required work was completed or that publication was possible.

Agent messages are omitted from this comment. The category uses structured run status, not interpretation of agent prose.

Generated from Ponytail Reviewer · routed: gpt56 max · ~2 AIC · ⌖ ~1 AIC · ⊞ 118 · ◷

Generated by Ponytail Reviewer for #74660

[github-actions]

Comment Memory

Peek at saved memory (pr-code-quality-reviewer)

reviewed_at: 2026-11-03T23:36:20Z
review_event: COMMENT
top_themes:
  - model-routing/sub-agent compatibility
files_reviewed:
  - .github/workflows/front-page-copy-guard.md
  - .github/workflows/purelock.md
  - .github/workflows/refactoring-cadence.md
comment_count: 1

Note

This comment is managed by comment memory.

Expand the saved memory block to view or edit the persistent context for this thread.
Edit only the text inside the backtick fences; workflow metadata and the footer are regenerated automatically.

Learn more about comment memory

🔎 Code quality review by PR Code Quality Reviewer · copilot · gpt50 · ~3 AIC · ⌖ ~1 AIC · ⊞ 1.6K · ◷
_{Comment /review to run again}

[github-actions]

Review complete. One actionable compatibility risk was found in the Purelock routing change; see the inline comment. The specialist pass completed, and speculative schema/test-evidence concerns were not reported.

🔎 Code quality review by PR Code Quality Reviewer · copilot · gpt50 · ~3 AIC · ⌖ ~1 AIC · ⊞ 1.6K
_{Comment /review to run again}

[Copilot]

🟢 Approval recommended

No active review issues are reported for the migration of three workflows to economy routing.

No open issues 🎉

What changed in this PR

Moves three gh-aw workflows from the retiring GPT-5.6 Terra pin to cost-oriented economy routing, addressing #74642 while preserving their existing tasks and permissions.

Changes:

• Enables GitHub-backed economy routing for two Codex workflows and one pi workflow.
• Regenerates execution locks with router setup, pinned images, and runtime model metadata.
• Removes inherited Terra pins from threat detection while leaving prompts, sub-agents, triggers, and safe outputs unchanged.

File	Description
.github/​workflows/​refactoring-cadence.md	Replaces the Terra pin with economy routing for pi.
.github/​workflows/​refactoring-cadence.lock.yml	Adds pi routing setup and propagates selected-model metadata.
.github/​workflows/​purelock.md	Enables Codex economy routing without changing sub-agents.
.github/​workflows/​purelock.lock.yml	Adds routing infrastructure and removes agent and detection Terra pins.
.github/​workflows/​front-page-copy-guard.md	Replaces the Terra pin with Codex economy routing.
.github/​workflows/​front-page-copy-guard.lock.yml	Adds routing setup, metadata reporting, and default detection-model selection.

_{Review effort: Balanced (auto)}

[github-actions]

Skills-Based Review 🧠

Applied /diagnosing-bugs to the routing change. Requesting changes because the economy router can still select the model this PR is meant to retire.

📋 Key Themes & Highlights

Key Themes

• Terra still selectable: model-routing with candidate-models defaulting to the gpt-* family (github-copilot/gpt-* in the compiled config) expands to every GPT model in the catalogue, including gpt-5.6-terra. None of the three workflows excludes it. Two (purelock, front-page-copy-guard) have no allowlist, and refactoring-cadence has none either. Routed smoke workflows already use allowed-models for this.
• Prompt-injection risk: The existing bot comment on purelock.lock.yml (line 1290) flags the routing pre-step, which writes the rendered prompt as a single user-role message. Worth confirming whether that risk is accepted.
• Compile warning: test-writer sub-agent is configured with the large model, which may be incompatible with the codex routing API. The author notes this as a follow-up.

Positive Highlights

• ✅ Clean, minimal frontmatter change; the removed model: pins are the right direction.
• ✅ Committed lock files reproduce from the HEAD sources, apart from scheduled cron comment lines.
• ✅ refactoring-cadence migration to the map form for engine is correct for pi.

Scope note: the local checkout is stale, so I reviewed committed HEAD and built gh-aw to recompile. I did not run the workflows, and I did not verify runtime ranking of Terra against Luna and Sol.

@copilot please address the inline comments above.

🧠 Reviewed using Matt Pocock's skills by Matt Pocock Skills Reviewer · copilot · routed: haiku55 xhigh · ~8 AIC · ⌖ ~1 AIC · ⊞ 5.4K
_{Comment /matt to run again}

[github-actions]

Warning

Threat detection engine error

Threat detection could not complete its analysis due to a tooling error. This is not a security finding.

What happened

The threat detection engine failed to produce results.

Review the workflow run logs to troubleshoot the error.

---

Note

The pull request was not created — a fallback review issue was created instead due to protected file changes: #74678

PR Finisher pass (submitted for publication, not confirmed pushed):

• Merged origin/main into the branch.
• Added allowed-models: [gpt-5.6-luna] to the economy routing in purelock, front-page-copy-guard and refactoring-cadence, which excludes Terra.
• Changed Purelock's test-writer sub-agent model from large to gpt-5.6-luna, which clears the compile warning.
• Compiled the three workflows. Lock files are excluded from the commit, so a maintainer needs to run make recompile. I did not run make agent-report-progress.
• The prompt-injection thread on routing pre-step system constraints is not addressed. It needs a change to the compiler or routing step, which is out of scope for this PR. Review threads are left open and no replies are staged, pending publication.
• CI is stale after the push and needs a maintainer re-trigger.

PR finisher dispatch requested for the following outstanding work: merge main, and the review items on purelock.lock.yml:1290, purelock.md:19, front-page-copy-guard.md:19 and refactoring-cadence.md:20.

Sous-chef head: 59b4e94
Sous-chef work: 07e9c5d026d68cf0faf8226253349f06ef3e48dbe47ea3bc55aee7bf14a1ec9e 3522992a14a1ea368f95d3f73815d784f02ec9d106d58a61b07b12eda967dc8f 368ad59d519944637a3570afce3cb2fae6dd6f8e39a96dad53e3cbe0fe123476 a47205db50b1b41061e8d4ae909144137744e0b33d08d03a45c8e89d5820bfed fb93ed6e385af43896670850a5bffebb16f8597afcbccf07cbb499779adbff76
Sous-chef state: 0fa0dcdc0815a68ea7ccf57e36736b1bfdf43dd633dc4c4f945aff7776216d2e

Generated by 🏁 PR Finisher · copilot · auto · ~23 AIC · ⊞ 10.7K · ◷

[github-actions]

---

Note

The pull request was not created — a fallback review issue was created instead due to protected file changes: #74704

PR Finisher

• Pinned test-writer sub-agent in purelock.md to gpt-5.6-sol (was large); the codex incompatibility compile warning is gone. Submitted for publication (lock file regenerated locally but excluded; needs recompile by maintainers). Review reply/resolve deferred until published.
• Could not merge main: local merge needs a network fetch (blocked, no credentials). Needs make merge-main by a maintainer.
• upload_code_coverage failure not investigated/fixed; likely needs maintainer re-run after push.

CI is stale after push and needs a maintainer re-trigger.

PR finisher dispatch requested for the following outstanding work (summary).

Sous-chef head: 59b4e94
Sous-chef work: 0fffc2b2c2895729dc76fc6b84b0435d1b53aaa1d5388f967ef50e6824edb91c fb93ed6e385af43896670850a5bffebb16f8597afcbccf07cbb499779adbff76
Sous-chef state: 392585207a6b12e89d00cd49d8f20b9e5d1ac2d875ec0857da5242cf45624456

Generated by 🏁 PR Finisher · copilot · auto · ~18 AIC · ⌖ ~1 AIC · ⊞ 10.7K · ◷

[github-actions]

PR Finisher: no changes submitted

• Merge main: blocked. The branch is 11 commits behind origin/main, and the local checkout lacks the objects to merge (promisor fetch fails without credentials). A maintainer should run make merge-main or use the Update branch button.
• purelock.md thread (test-writer model: large): left open. A collaborator already replied that routing is intended. Whether routing is compatible with the sub-agent large class still needs a maintainer decision or a dry run. I could not verify either here.
• upload_code_coverage failure: logs were unreadable from this run (403), so no fix was made. Re-check after updating the branch.
• Terra allowlist threads: already answered by a collaborator as intentional; no changes.

CI will need a maintainer re-trigger after the merge.

Sous-chef head: 59b4e94
Sous-chef work: 0fffc2b2c2895729dc76fc6b84b0435d1b53aaa1d5388f967ef50e6824edb91c fb93ed6e385af43896670850a5bffebb16f8597afcbccf07cbb499779adbff76
Sous-chef state: 392585207a6b12e89d00cd49d8f20b9e5d1ac2d875ec0857da5242cf45624456

Generated by 🏁 PR Finisher · copilot · auto · ~16 AIC · ⌖ ~1 AIC · ⊞ 10.7K · ◷

Addressed: Terra drops out of the live pool when it retires, and #74642 asks for no allowed-models list. test-writer was verified with routed run 41745095607 (all requests 200). Threads answered and resolved.

[github-actions]

🎉 This pull request is included in a new release.

Release: v1.2.0