normalize-fetch-execution

name	normalize-fetch-execution
description	Execute the current actor's slice of one prepared fetch-plan through explicit queue-runner, normalizer-runner, and execution-receipt components. Use when an investigator needs their assigned raw artifacts copied/fetched, normalized into DB signal rows, and recorded without selecting downstream analysis conclusions.

Eco Import Fetch Execution

Core Goal

• Read the prepared fetch plan for the current round.
• Resolve the executing actor role from --actor-role or the runtime-injected OPENCLAW_ACTOR_ROLE.
• Run only fetch-plan steps owned by that actor's role, for example social-investigator for social_investigator steps and environmental-investigator for environmental_investigator steps.
• Run the queue_runner component for owned steps: copy local imports or execute approved detached-fetch steps into the current run raw store.
• Run the normalizer_runner component for owned steps: invoke mapped normalizer skills or keep a raw-only receipt when no normalizer exists.
• Run the execution_receipt component: write one auditable execution snapshot with step status, raw artifact refs, normalizer receipts, and warnings.
• Do not choose claim extraction, observation extraction, coverage scoring, readiness, report basis, or any other analysis chain.

Triggering Conditions

• prepare-round already wrote fetch_plan_<round_id>.json.
• The current ingress path may contain both local artifact imports and detached fetch requests.
• Each investigator runs their own role slice; runtime-operator approves/governs but does not execute data fetch or normalization.
• Downstream extraction should start from normalized signal-plane data rather than direct seed helpers.

Read/Write Contract

• Reads run_dir/runtime/fetch_plan_<round_id>.json.
• Writes run_dir/raw/<round_id>.
• Writes run_dir/analytics/signal_plane.sqlite.
• Writes run_dir/runtime/import_execution_<round_id>.json.
• Output board_handoff.suggested_next_skills is limited to DB query surfaces.

Required Input

• run_dir
• run_id
• round_id
• actor_role through --actor-role or runtime OPENCLAW_ACTOR_ROLE

Output Contract

• status
• summary
• receipt_id
• batch_id
• artifact_refs
• canonical_ids
• warnings
• board_handoff
• execution_components

References

• ../../docs/openclaw-project-overview.md
• ../../docs/openclaw-skills-refactor-checklist-v2.md

Scripts

• scripts/normalize_fetch_execution.py