causal-consultant

star 4

Interactive consultation-style causal inference skill. Use when the user wants to work with an agent on a causal question, causal data analysis, causal design and discovery, method choice, interpretation, diagnostics, or report writing. The skill interacts with the user to turn a raw idea such as 'analyze the causal effect of X on Y' into a defensible causal question before running or recommending analysis.

rqzhu-aide By rqzhu-aide schedule Updated 6/6/2026

name: causal-consultant description: Explicit-use interactive causal consulting team. Use only when the user explicitly asks to use causal-consultant, the causal consultant skill/team, or an interactive persistent causal-consultant workflow for project intake, data audit, domain expertise, causal discovery, causal checking, report writing, methods/results interpretation, reviewer-response support, or project state tracking.

Causal Consultant Router

This file is only the router. It initializes state, prepares next_step_plan, and then loads the planned reference files. Manager synthesis, state-file checks, and final user-facing responses belong to references/team_lead.md.

Turn Protocol

  1. Resolve the project root. If unclear, use the current working directory.
  2. Run scripts/init_project_state.py --project-root <project root> to initialize project_state.yaml if needed.
  3. Read project_state.yaml.
  4. Read references/route_index.yaml.
  5. Read references/route_selection_workflow.md. After reading the current user message, always run route selection before any substantive causal work, writing, file inspection, or output creation. The router creates next_step_plan; it does not satisfy the user request directly.
  6. Follow references/route_selection_workflow.md for route selection. Load conditional routing references only when that file says the current turn needs them.
  7. Write the complete next_step_plan assignment list to project_state.yaml before loading any route reference.
  8. Load planned non-team_lead references first. Core routes read their own next_step_plan entry. For analysis_execution, load the named design and optional support references; there is no separate analysis_execution reference file.
  9. After non-team_lead route work is done or blocked, load references/team_lead.md exactly once as the final planned reference. Let team_lead load conditional report, analysis, or artifact support only when its review needs them.
  10. Let team_lead review the prepared plan and route-owned work, update aggregate YAML fields, check the state file, and produce the only final user-facing answer.

Keep route selection, file reads, and file writes silent unless there is a blocker or permission issue.

Install via CLI
npx skills add https://github.com/rqzhu-aide/causal-consultant --skill causal-consultant
Repository Details
star Stars 4
call_split Forks 1
navigation Branch main
article Path SKILL.md
More from Creator