[xnoto]

Summary

Use openai/gpt-6.1-terra with the existing medium variant for only the generic terra execution subagent. All ten primaries, the global Sol fallback, and all five Sol-backed reviewers remain unchanged. Bump the immutable opencode-server chart from 0.5.6 to 0.5.7; appVersion 2.0.22 and API selector 2 are unchanged.

This follows the owner’s revised subagents-only design goal. The branch history includes and then explicitly undoes the initial primary/reviewer changes; the final net runtime diff is one model line.

Evaluate roles rather than blindly restoring history: keep architecture, DevOps, hostile-code, infrastructure-security, and recruiter review on Sol. All six former Sol-backed subagents were historically Terra; no newer Sol-only subagent was found. Small paired synthetic coding/resume responses inform this conservative first pass, not broad quality or token-efficiency claims. Terra missed an evidence distinction in the resume sample that Sol identified.

OpenAI lists standard input/output rates 50% below Sol (Terra, Sol). This is not a claim about account-plan billing, fewer tokens, or complete-task savings.

Fixes: N/A — owner-requested model-policy refinement.

Type of change

• Feature / enhancement
• Documentation
• GitOps desired state (manifests, kustomize, charts, SOPS/KSOPS secrets)
• CI / reusable workflow

Validation

• Required pull-request checks pass — CI pending when opened; no local checks were run.
• Generated or centrally distributed files were regenerated by their owning automation, not hand-edited — canonical packaged inputs only; Helm renders ConfigMap/mount/checksum data in CI.

Expected checks: repository hygiene, strict Helm lint/render, exact agent inventory and model/variant checks, archive byte parity, and memory-pilot Python tests. Keep both historical baselines and existing transforms unchanged; add one exact Terra-only header transform with malformed/duplicate-selection and unchanged-body/unrelated-file fixtures. The independent eighteen-agent source/render matrix changes only Terra; ten-primary Sol checks remain intact.

Pre-PR reviews: adversarial ADVANCE (one Low wording correction applied), infrastructure-security ADVANCE for the final two-file runtime diff, DevOps integration ADVANCE. QA mapped static coverage and identified later runtime-engagement/consumer gates; release review found version/pin consistency, with its then-pending independent reviews now completed. No Critical/High finding remains. These are independent agent reviews, not GitHub approval or CI results.

Catalog lookup confirms the exact Terra ID and medium variant. Paired synthetic responses exercised supplied text only: code was not executed, and deployed tool-call behavior was not tested. Before accepting rollout, define a representative task rubric and verify fresh child model/variant, inference/tool calls, factual fidelity, correction burden, and usage/latency; separately verify unchanged primary Sol selection.

Impact and rollout

Producer: makeitworkcloud/charts, main target, chart opencode-server:0.5.7. Net paths: Makefile, opencode-server/Chart.yaml, files/agents/terra.md, tests/test_memory_pilot_render.py, chart README, and the architecture/memory-pilot guides. No other runtime model, prompt, permission, provider, image, credential, template, or pilot change.

• Authored/reviewed: this branch. PR validation and packaging: automatic after creation.
• Merge/publication: confirmation-gated; not authorized or performed. A confirmed merge triggers the existing main workflow to publish the immutable OCI chart to GHCR.
• Consumer selection: after successful publication, existing automation opens/updates an auto-merge pin PR in makeitworkcloud/kustomize-cluster; required checks gate it. At 00469e89, the Application and coupled assertion still select 0.5.6/API 2. Update the pin and exact assertion atomically to 0.5.7/2 in that future consumer PR. Do not merge an assertion-only change or bypass checks. No consumer edit is included here.
• Reconciliation/health: not performed. Future pin merge can trigger automated Argo reconciliation and a Recreate pod replacement via ConfigMap checksum; config loads at startup, not hot reload. Verify root gitops-workloads, child opencode, rollout, health, and function separately.
• Rollback: separately confirmed GitOps pin reversion to the prior published 0.5.6 chart.

Safety and secrets

• Contains no plaintext secrets, decrypted SOPS values, state files, kubeconfigs, tokens, or private endpoints.
• No local OpenTofu init/plan/apply/destroy/import/state operations were run or claimed — plans come from pull-request checks.
• Breaking or irreversible effects are described above with rollback notes.

AI-assisted implementation, bounded synthetic evaluation, and independent review. No merge, publication, workflow dispatch, or live mutation is authorized by this PR.

[github-actions]

Chart CI passed

Repository hygiene and Helm validation passed. View the workflow run.