identify

name	identify
description	Design identification strategy by dispatching the causal-strategist agent (proposer) and identification-critic agent (critic). Produces strategy memo, pseudo-code, robustness plan, and falsification tests. Use when asked to "identify the effect", "design the strategy", or "write the strategy memo".
disable-model-invocation	true
argument-hint	[research question or research-spec file path]
allowed-tools	["Read","Grep","Glob","Write","Task"]

Identify

Design an identification strategy by dispatching the causal-strategist (proposer) and identification-critic (critic).

Input: $ARGUMENTS — research question, description of available variation, or path to research spec.

---

Workflow

Step 1: Context Gathering

1. Read research spec if exists (quality_reports/research_spec_*.md)
2. Read literature review if exists (quality_reports/lit_review_*.md)
3. Read data assessment if exists (quality_reports/data_exploration_*.md)
4. Read .claude/rules/domain-profile.md for common identification strategies in the field

Step 2: Launch causal-strategist Agent

Delegate to the causal-strategist agent via Task tool:

Prompt: Design identification strategy for "[research question]".
Available data: [from context gathering].
Available variation: [from research spec or user description].

Produce:
  1. Strategy memo — design choice, estimand, assumptions, comparison group
  2. Pseudo-code — implementation sketch (what the R/Stata code will do)
  3. Robustness plan — ordered list of robustness checks with rationale
  4. Falsification tests — what SHOULD NOT show effects (and why)
  5. Referee objection anticipation — top 5 objections with pre-emptive responses

If multiple credible strategies exist, present the top 2-3 with trade-offs.
Save strategy memo to quality_reports/strategy_memo_[topic].md

Step 3: Launch identification-critic Agent (Strategy Review Mode)

After causal-strategist returns, delegate to the identification-critic agent:

Prompt: Review the strategy memo at quality_reports/strategy_memo_[topic].md.
Mode: Strategy review (Phase 2 focus — core design validity).
Check:
  - Are the identifying assumptions stated and defensible?
  - Is the estimand clearly defined (ATT/ATE/LATE)?
  - Are the biggest threats acknowledged?
  - Does the robustness plan address the right concerns?
  - Are the falsification tests well-chosen?
Score the strategy. Flag CRITICAL issues that must be resolved before coding.
Save review to quality_reports/strategy_memo_[topic]_identification_review.md

Step 4: Iterate if Needed

If identification-critic finds CRITICAL issues:

1. Present issues to user
2. Re-dispatch causal-strategist with specific fixes (max 3 rounds per three-strikes.md)
3. Re-run identification-critic to verify

If unresolved after 3 rounds: escalate to user with both perspectives.

Step 5: Present Results

# Identification Strategy: [Topic]
**Date:** [YYYY-MM-DD]
**Design:** [DiD / IV / RDD / SC / Event Study]
**Estimand:** [ATT / ATE / LATE]

## Strategy Summary
[2-3 sentence description of the identification strategy]

## Key Assumptions
1. [Assumption 1] — [how it will be defended]
2. [Assumption 2] — [how it will be defended]

## identification-critic Assessment: [SOUND / CONCERNS / CRITICAL ISSUES]
- Critical issues: N
- Major issues: N
- Strategy memo score: XX/100

## Robustness Plan (ordered)
1. [Most important check]
2. [Second check]
...

## Next Steps
- [ ] Implement main specification (→ /data-analysis)
- [ ] Run falsification tests
- [ ] Generate pre-trend evidence / first stage / McCrary

---

Principles

• causal-strategist proposes, identification-critic critiques. The adversarial pairing catches design flaws early.
• Strategy memo is the contract. Once approved, the Coder implements it faithfully.
• Catch problems before coding. A flawed strategy caught now saves weeks of wasted analysis.
• Multiple strategies are OK. Present trade-offs and let the user choose.
• The user decides. If causal-strategist and identification-critic disagree after 3 rounds, the user resolves it.