docs: reroute agent-loop models by task, add frontier-first reviewer cascade - #101
Merged
Conversation
Splits area:docs/area:tests into separate model tiers (were sharing MODEL_CHEAP), documents 2-3 fallback models per tier based on 2026-08 Cursor Pro / OpenCode Go catalog research, and gives the reviewer a real frontier-first cascade (gpt-5.3-codex -> gpt-5.6-sol via Cursor, falling back to kimi-k3 via OpenCode) instead of always running on OpenCode. Also fixes a stale README claim (said gpt-5.3-codex was the frontend model; the script has always used composer-2.5) and documents the model routing table in both the agent runbook and the top-level README. Several model ids (gpt-5.3-codex, gpt-5.6-sol, glm-5.2, qwen3.7-max, deepseek-v4-flash, mimo-v2.5) are sourced from web research and not yet verified against a live catalog call -- flagged in the docs, sanity-check with workflow_dispatch on a low-stakes issue before trusting the cron. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Ekv1VKFK4VVEdcPSGhA7x
github-actions
Bot
added
area: ci-cd
This was referenced Aug 24, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Descripción
Reasigna los modelos del
agent-looppor tarea, según investigación de catálogo/precios de Cursor Pro y OpenCode Go (2026-08):gpt-5.3-codex→gpt-5.6-sol(Cursor, pool "Other Models") →opencode-go/kimi-k3como respaldo garantizado (antes: siempre OpenCode).kimi-k3, documentadas alternativasglm-5.2/qwen3.7-max.area: docsyarea: testsdejan de compartirMODEL_CHEAP: cada una tiene su propia cadena de fallback (tests escala aglm-5.2en lógica compleja; docs se queda en el tier más barato).composer-2.5), documentadas alternativasgrok-4.5/grok-4.6..github/agent/README.md(decíagpt-5.3-codexpara frontend; el código siempre usócomposer-2.5).Varios ids (
gpt-5.3-codex,gpt-5.6-sol,glm-5.2,qwen3.7-max,deepseek-v4-flash,mimo-v2.5) vienen de investigación web, no verificados contracursor-agent --list-models/ el catálogo real de OpenCode Go. Recomendado: probar congh workflow run agent-loop -f issue=<N>en un issue de bajo riesgo antes de confiar en el cron automático (3 runs/día).Checklist
bash -nsobrepick-issue.sh(sintaxis OK)agent-loop.yml🤖 Generated with Claude Code
https://claude.ai/code/session_012Ekv1VKFK4VVEdcPSGhA7x