Declared configuration and Blueprint

What changed?

Standalone Playbook authoring, cloning, purchasing, and Lab simulation have retired. Badge preserves existing rows, including their immutable versioning with parent chains, but they no longer accept customer mutations and they are not evidence of what an agent currently runs.

Declared configuration

Owners and workspace editors can see a narrow, read-only projection of historical declarations in the agent's Evidence section. It may include a declared model, temperature, token limit, sampling value, type, version, and active/default status.

This projection deliberately excludes system prompts, examples, tools, notes, pricing, internal identifiers, and other private fields. It is labelled:

Self-declared configuration — Badge does not apply or verify these values, and they may not match the current runtime.

Anonymous visitors and non-editor viewers do not receive this projection.

Blueprint is observed evidence

Blueprint is built from telemetry Badge actually receives from connected agent runs. It belongs in the same Evidence section so you can compare what was once declared with what Badge has observed, while keeping the two sources visibly separate.

Badge does not automatically claim that a declaration matches or conflicts with Blueprint. Your live endpoint remains authoritative for runtime configuration, and changing a preserved declaration would not change endpoint behaviour.

Evaluate changes with Screen

Change the configuration on infrastructure you control, then use Screen to produce a real evaluation. Open the agent's Evidence, Runs, and Improve sections to inspect the resulting telemetry, scores, and diagnostics.