Skip to content

Agent Model Config

.github/agent-config.yaml is the single source of truth for which Claude model backs each agentic workflow. Before this config, the model was hardcoded (and inconsistently) in ~8 workflow files, so a repo-wide model bump meant hand editing each one. Now it is one edit (issue #5218).

How it works

Each agent workflow resolves its model at run time via the composite action .github/actions/resolve-agent-config. The action reads agent-config.yaml, resolves the effective model, and exports it as the AGENT_MODEL environment variable for the following steps, which pass it to claude-code-action as --model ${{ env.AGENT_MODEL }}.

Resolution order (highest priority first):

  1. an explicit per-run override — a workflow_dispatch model: input, passed to the action as model-override;
  2. the per-workflow model: in agent-config.yaml;
  3. the top-level default_model.

So scheduled runs use the config; a manual dispatch may still pick a different model from the dropdown (that selection overrides the config for that run only).

Changing a model

To change the model a workflow uses on its scheduled runs, edit its model: in .github/agent-config.yaml — no workflow YAML changes needed. To bump every workflow that has no explicit model:, change default_model.

# .github/agent-config.yaml
default_model: claude-opus-5
workflows:
  literature-scan:
    model: claude-haiku-4-5-20251001
  pr-shepherd:
    model: claude-opus-5
  # ...

The curation-scanner matrix

curation-scanner fans out across effort tiers in parallel, each tier bound to a model and a GitHub label selector. Its full fan-out lives in agent-config.yaml under a matrix: list; a setup job reads it with the resolver's --matrix mode and feeds it to strategy.matrix.include:

workflows:
  curation-scanner:
    matrix:
      - { effort: low_effort,    model: claude-haiku-4-5-20251001, selector: "..." }
      - { effort: medium_effort, model: claude-sonnet-5,         selector: "..." }
      - { effort: high_effort,   model: claude-opus-5,           selector: "..." }

Adding a workflow

  1. Add a workflows.<stem> entry with a model: (or a matrix:).
  2. In the workflow, after checkout, add: 
    - name: Resolve agent config
  uses: ./.github/actions/resolve-agent-config
  with:
    workflow: <stem>
    model-override: ${{ inputs.model }}   # omit if no model input
  3. Use --model ${{ env.AGENT_MODEL }} in the agent invocation.

tests/test_agent_config.py enforces the contract: every workflows: key maps to a real workflow file that invokes the action, and no workflow hardcodes a --model claude-* inline.

Scope and relationship to cron-profiles

This config covers the AI model only. Cron cadence stays in .github/cron-profiles.yaml because GitHub cannot parameterize schedule: at run time. Per-workflow enable/disable and a provider abstraction (provider: is reserved but not yet consumed) are planned follow-ups under issue #5218.

claude.yml (the @claude mention responder) intentionally leaves the model unset and uses claude-code-action's own default, so it is not managed here.