spec-kitty-runtime

Canonical mission runtime for Spec Kitty.

Scope (V1)

1. YAML mission template loading.
2. Deterministic mission discovery with precedence tiers.
3. Deterministic step planner and mission next() execution loop.
4. Prompt rendering for agent runtimes.

Public API

1. load_mission_template(path_or_key, context=None)
2. discover_missions(context)
3. start_mission_run(template_key, inputs, policy_snapshot, context=None, run_store=None)
4. next_step(run_ref, agent_id, result="success", policy_snapshot=None, actor_context=None, context=None)
5. render_prompt(decision, format="markdown")

Runtime Event Emitter API (v0.2.0a0)

RuntimeEventEmitter uses canonical payload-object methods aligned to spec-kitty-events mission-next contracts:

1. emit_mission_run_started(payload)
2. emit_next_step_issued(payload)
3. emit_next_step_auto_completed(payload)
4. emit_decision_input_requested(payload)
5. emit_decision_input_answered(payload)
6. emit_mission_run_completed(payload)

NullEmitter is the default no-op implementation for local-first/offline flows.

Step Context Contracts (V1)

Mission steps declare their context requirements using step context contracts:

• requires: Contexts that MUST resolve before execution
• optional: Contexts that may enrich but aren't blocking
• emits: Contexts produced/updated on completion

See docs/contracts.md for the complete reference, V1 baseline context types, and integration guide for spec-kitty 2.x.

Mission Pack Layout (YAML-only)

mission-pack.yaml
missions/<mission_key>/mission.yaml
missions/<mission_key>/templates/*.md
missions/<mission_key>/steps/*.yaml  # optional