Repository navigation
feat(pstack): replace Luna workers and upgrade reflection - #25
Conversation
There was a problem hiding this comment.
Copilot review overview
🟢 Approval recommended
The changes are internally consistent documentation/config routing updates verified aligned across every mirrored copy (config, template, routed skills, validator, executable defaults, and changelog) with no logic changes.
Review effort: Balanced
Findings: None
What changed in this PR
This PR completes the migration of PStack's per-role model routing to the GPT-6.1 Sol family. It replaces the remaining active Luna routes (discovery, investigation, Swarm workers, comment audits, and live-verification lanes) with gpt-6.1-sol at low, promotes the hardest-tasks and Reflect judgment/divergent/synthesis roles to gpt-6.1-sol at high, and corrects a lingering Autoreview prose summary that still cited "Astra medium." The changes keep the routing config (models.md), the setup template, the routed skill fallbacks, the plan validator, and the changelog all in sync, following up on the review comment from #24.
Changes:
- Replace Luna discovery/investigation/Swarm/comment-audit/live-verification routes with
gpt-6.1-solatlow. - Promote hardest-tasks and Reflect judgment/divergent/synthesizer roles to
gpt-6.1-solathigh. - Fix the Autoreview summary to
gpt-6.1-solatxhigh, matching its executable and documented defaults.
| File | Description |
|---|---|
| skills/poteto-mode/references/models.md | Updates the canonical per-role defaults for hardest tasks, how/why explorers, reflect roles, swarm, and Comment Sicko. |
| skills/setup-pstack/SKILL.md | Mirrors the same role defaults in the setup template. |
| skills/reflect/SKILL.md | Moves reflect judgment/divergent/synthesizer defaults to gpt-6.1-sol high. |
| skills/how/SKILL.md | Sets the how-explorer fallback to gpt-6.1-sol low. |
| skills/why/SKILL.md | Sets the why-investigators fallback to gpt-6.1-sol low. |
| skills/swarm/SKILL.md | Sets the swarm-worker fallback to gpt-6.1-sol low. |
| skills/no-comments/SKILL.md | Sets the Comment Sicko fallback to gpt-6.1-sol low. |
| skills/poteto-mode/SKILL.md | Points the "strongest judgment model" for hardest changes at gpt-6.1-sol high. |
| skills/poteto-mode/scripts/check-plan.mjs | Updates the LANES validator constant to the new live-lane model wording. |
| skills/poteto-mode/playbooks/multi-phase-plan.md | Updates live-verification lane wording to gpt-6.1-sol low to match the validator. |
| skills/autoreview/SKILL.md | Corrects the prose summary from Astra medium to gpt-6.1-sol xhigh. |
| CHANGELOG.md | Records the routing changes under the 2026-09-29 heading. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Replace active Luna routes with GPT-6.1 Sol low for discovery, investigation, Swarm workers, comment audits, and live verification. Move hardest tasks and Reflect judgment/divergent/synthesis to GPT-6.1 Sol high. Align role configuration, setup templates, skill fallback defaults, and the plan validator.
Fix the remaining Autoreview summary that still said Astra medium, addressing the review comment on #24; the summary now matches the Sol 6.1 xhigh executable default.
Validation: