co-scientist-run

co-scientist-run

Goal:

• Start one Co-Scientist run from Claude Code and then continue with the canonical pipeline.

Expected input:

• one run directory such as runs/test1
• or one compatibility config path such as runs/test1/config.yaml

Execution steps:

1. Resolve the run directory or compatibility config path relative to the repository root when the user gives a relative path.

2. Run:

   python -m tools.host.claude_project_cli run <run-target> --skill co-scientist-pipeline
   

3. Read the emitted handoff artifact:

   • runs/<run_id>/state/HOST_AGENT_HANDOFF.json

4. Read the CLI JSON result and the run-local dashboard receipt artifacts:

   • runs/<run_id>/dashboard/LINKS.md
   • runs/<run_id>/dashboard/LINKS.json

5. If the CLI JSON contains dashboardLinks:

   • If dashboard.status is running, return dashboardLinks.dashboard as the primary dashboard URL and include the deep links.
   • If dashboard.status is starting, immediately run:

   python -m tools.host.claude_project_cli dashboard <run-dir>
   

   • Read the refreshed CLI JSON result plus runs/<run_id>/dashboard/LINKS.md.
   • If runtime.status is now running, return the refreshed links.dashboard URL as the primary dashboard URL and include the deep links.
   • If runtime.status is still starting, tell the user that the dashboard is still booting, point them to runs/<run_id>/dashboard/LINKS.md, and include the retry command:

   /co-scientist-dashboard <run-dir>
   

6. Open the canonical workflow and shared contracts:

   • skills/co-scientist-pipeline/SKILL.md
   • skills/shared-references/artifact-contract.md
   • skills/shared-references/state-contract.md
   • skills/shared-references/integration-contract.md
   • skills/shared-references/execution-modes.md
   • skills/shared-references/schema-index.md

7. Continue execution from the canonical repository-local skills tree instead of inventing a parallel flow. If the refreshed routing plan returns run_configuration, execute research-config and validate research_plan/RESEARCH_PLAN.json before any generation work.

8. After each major phase write, run:

   python -m tools.validation.contract_validation runs/<run_id> --skill co-scientist-pipeline
   

Rules:

• Read execution semantics from repository-local SKILL.md files and dynamic context from canonical artifacts.
• Use runs/<run_id>/state/PIPELINE_STATE.json and CURRENT_STAGE.json as the authoritative resume state.
• The first active stage may be Configuration rather than Generation; do not skip it when the routing plan requires run_configuration.
• If the effective run policy is iteration_policy = completion_driven and human_checkpoint = auto, do not ask whether to continue after each evolution round. Keep running until a real terminal route, checkpoint boundary, or blocking validator/safety state is reached.
• If you must stop before convergence or a terminal route, tell the user the run is paused, current convergence has not been reached, persisted state is resumable, and the next recommended action is continue evolution through resume or an explicit continue request.
• Treat skills/ as canonical and .claude/skills/ as the Claude Code entry surface.
• Treat runs/<run_id>/dashboard/LINKS.md as the human-readable dashboard receipt and runs/<run_id>/dashboard/LINKS.json as the machine-readable receipt.