Component · spec & evidence

Provider Context Recipe Budget Policy

Runs the real context harness to measure assembled byte sizes and check each bundle fits its budget.

Contract

Does This component checks that the bundles of context an AI agent would assemble before calling an outside model provider stay inside fixed size limits (in bytes), fill their sections in the declared order until the budget runs out, list any section that was dropped for not fitting, and never carry answer keys, proof solutions, or other "correct answer" material. The record shows the exact size ceilings for each recipe, which sections fit versus got left out, and which output each recipe is allowed to produce, so the context boundary is inspectable as plain accounting before any external model access or answer authority is ever in play. It only validates this metadata; it does not itself call any provider.

Scope limit It validates context-budget projection mechanics (byte ceilings, ordered section fill, omitted-section manifests, deliverable routing, and digest-checked source-body imports) only. It excludes provider/API calls, run Lean/Lake, expose or carry proof or oracle truth-side material, assert theorem or domain-level conclusions, or include launch operations.

Run

PYTHONPATH=src python3 -m microcosm_core.organs.provider_context_recipe_budget_policy run --input fixtures/first_wave/provider_context_recipe_budget_policy/input --out receipts/first_wave/provider_context_recipe_budget_policy

Evidence

ClassBounded runtime computationevidence 4/5Real runtime resultRuns real tools

Links to Bounded Autonomy Campaign Packet, Tool Server Pressure Inventory

ai-safetyagent-evaluationred-teaming

Source

Source Design note · Source

Map Show on the map

Index Open compact index card

Paper module