propose-next-actions

name	propose-next-actions
description	Optional moderator advisory skill for proposing possible next actions from DB-backed council state. It requires operator-approved skill approval and never owns the default investigation sequence.

Eco Propose Next Actions

Core Goal

• Suggest possible follow-up actions when a moderator asks for advisory help.
• Label ranking logic as heuristic and keep it subordinate to DB-native findings, evidence bundles, proposals, and moderator judgement.
• Emit a durable optional-analysis artifact that can be audited, consumed, or ignored without changing phase state.
• Require request-skill-approval -> approve-skill-approval -> run-skill --skill-approval-request-id before execution.

Triggering Conditions

• A moderator has identified a planning question that would benefit from optional ranking support.
• The operator has approved this optional-analysis run for the specific round and actor role.
• Need advisory candidates for future investigation, not a required queue before probes, readiness, report basis, or reporting.

Read/Write Contract

• Syncs the round into the run-local deliberation plane and prefers that state for action ranking.
• Reads run_dir/board/board_state_summary_<round_id>.json by default as a compatible advisory fallback.
• Reads run_dir/board/board_brief_<round_id>.md by default when present.
• Reads evidence coverage from the run-local analysis plane first.
• Falls back to run_dir/analytics/evidence_coverage_<round_id>.json when the synced result set is unavailable.
• Writes run_dir/investigation/next_actions_<round_id>.json as an optional advisory export and syncs canonical rows where supported.

Required Input

• run_dir
• run_id
• round_id
• Optional:
  • board_summary_path
  • board_brief_path
  • coverage_path
  • output_path
  • max_actions

Output Contract

• status
• summary
• receipt_id
• batch_id
• artifact_refs
• canonical_ids
• warnings
• deliberation_sync
• analysis_sync
• board_handoff
• The emitted artifact also carries normalized D1 trace metadata in board_state_source, coverage_source, db_path, and observed_inputs, including explicit *_artifact_present and *_present flags.

References

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

Scripts

• scripts/propose_next_actions.py