Skip to content

A real system prompt and prompt caching #12

Open

Description

@agent3-workflow

Do.

  • System prompt: cwd, OS, shell, date, git branch/status/recent commits, tool guidance. Byte-stable across turns.
  • Anthropic cache_control on system, tools and the last messages; OpenAI prompt_cache_key = session id. Report cache tokens in usage.
  • Drop hard-coded temperature: 0.1 for models that reject it (OpenAI reasoning models); make max_tokens per model.

Spec §5.1 #7–8.

Spec: SPEC_MUXCODE_HARNESS_AND_FEATURE_GAPS_2026_09_26.md

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    P1highenhancementNew feature or requestphase-1core agent loop

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions