Campaign Result Contract

Campaign execution writes one primary self-contained result:

runs/campaign/<campaign-id>/campaign_plan_result.json

The current schema is campaign_plan_result_v5.

Required Sections

Field	Meaning
schema_version	Result contract version
campaign_id	Campaign identifier when present in the request
status	Top-level planner outcome
selected_strategy	Strategy selected for the result
summary	Headline assignments, events, duration, propellant, and closure
request_snapshot	Canonical request that was evaluated
selection	Strategy/refinement choice and feasibility signals
campaign_summary	Assignment, event, mission-burden, objective, and target summaries
reference_policy	Reference and comparison policy recorded by the planner
assignments	Satellite-to-asset, plane, and slot assignments
events	Ordered campaign event sequence
missions	Mission-level outcomes on the selected path
diagnostics	Decision-path and execution diagnostics

request_snapshot is required. It makes the result reviewable without the original authoring file and is the correct input for comparison and provenance.

Status Model

Top-level status accepts:

• success
• partial
• infeasible
• failed

campaign_summary.status_class provides the engineering review class:

• success
• partial_success
• infeasible
• degraded
• failed

The process exit status and the campaign engineering status answer different questions. A successful execution can correctly return a partial or infeasible campaign.

Review Order

1. Read status and campaign_summary.status_class.
2. Check target closure and residuals in campaign_summary.target_summary.
3. Inspect assignments and overrides.
4. Review event sequencing and total burden.
5. Drill into non-success missions and reason codes.
6. Use selection and diagnostics to understand why the path was chosen.

Mission solve reuse, when enabled, is reported through aggregate cache counters and diagnostics. A reused exact-signature result records provenance while the campaign contract and source mission outcome remain available for review.

Optional Mission Bundles

With mission_artifact_detail: full, Astraeus may also write:

campaign_plan_result.missions/
|-- campaign_mission_bundles_manifest.json
`-- <satellite-id>/
    |-- mission_plan_request.json
    `-- mission_plan_result.json

These side bundles support targeted drilldown; campaign_plan_result.json remains the campaign authority.

Deterministic Review

astraeus review campaign --result <campaign_plan_result.json>

The review is a deterministic reader of the saved result and writes a compact markdown assessment while leaving campaign execution unchanged.

When comparing variants, compare the embedded request snapshot, outcome class, target closure, mission burden, and sequencing—not only total duration or propellant.