| name | strategize |
| description | Design identification strategy or pre-analysis plans using Strategist plus strategist-critic. Use when the user asks for a causal design, robustness plan, falsification tests, estimator choice, or a PAP. |
Strategize
Design an identification strategy or pre-analysis plan by dispatching the Strategist (proposer) and strategist-critic (validator).
Read First
AGENTS.md
plugins/clo-author/references/runtime-activation.md
plugins/clo-author/references/codex-workflow.md
plugins/clo-author/references/agent-catalog.md
plugins/clo-author/references/critic-runtime-contract.md
plugins/clo-author/references/domain-profile.md
plugins/clo-author/references/logging.md
templates/decision-record.md when a major strategy choice is being locked
Modes
- Identification strategy
- Pre-analysis plan
- Interactive PAP interview
Identification Strategy Workflow
Step 1: Gather Inputs
- Read the research spec, literature review, and data assessment if they exist.
- Read
plugins/clo-author/references/domain-profile.md for common identification patterns and referee concerns in the field.
- If key requirements are still unclear, create a spec before proceeding.
Step 2: Produce a Pre-Strategy Report
Before proposing a design, produce a Pre-Strategy Report that states:
- which research spec, literature review, and data assessment files were found
- whether the domain profile was loaded
- the current research question in one sentence
- the key findings from the literature and the data variation available
- any missing inputs that will be carried forward as explicit
[ASSUMED] items
Step 3: Dispatch Strategist
The Strategist should produce:
- a strategy memo with design choice, estimand, treatment, comparison group, and assumptions
- pseudo-code or implementation sketch for the main estimation
- a robustness plan ordered by priority
- falsification tests that should not show effects
- top referee objections with proposed responses
The strategy memo should cover:
- estimand
- estimator
- identifying assumptions
- threats to inference
- robustness plan
- falsification tests
Step 4: Dispatch strategist-critic
Run strategist-critic review through plugins/clo-author/references/active-critics/strategist-critic.md before treating the design as approved. The strategist-critic should review through the preserved four-phase protocol:
- Claim identification
- Core design validity
- Inference soundness
- Polish and completeness
The review must produce:
- a saved review report under
quality_reports/
- severity-labelled issues
- a score
- an approval state for the research journal
Step 5: Revision Loop
If CRITICAL issues found, iterate (max 3 rounds per three-strikes rule).
- Round 1: strategist revises from the critic's blocking issues
- Round 2: strategist-critic re-reviews the revised memo
- Round 3: if still blocked, escalate to the user because the disagreement is strategic rather than editorial
The design is not approved until the saved strategist-critic artefact clears the active gate.
Step 6: Save Outputs
Save to:
quality_reports/strategy_memo_[topic].md
quality_reports/strategy_memo_[topic]_review.md
quality_reports/decisions/strategy_[topic].md when a design choice is locked
Append a research-journal entry with round number, score, approval, report path, and next actor.
PAP Workflow
When the user asks for a PAP, adapt the strategy output into the requested registration format.
Interactive PAP Interview
For an interactive PAP flow, ask for:
- the research question
- the study design
- the primary outcomes
- the source of identifying variation or assignment mechanism
- pre-specified subgroup analyses
- the multiple-testing plan
PAP Requirements
The PAP should specify:
- study overview
- outcomes
- estimating equations
- subgroup analyses
- multiple-testing correction
- power assumptions when relevant
- sample and exclusion rules
- data and analysis plan
- timeline
- deviations log template
Mark every unresolved item with [ASSUMED], and end with a review checklist so the user can confirm assumptions before registration.
Principles
- Strategist proposes, strategist-critic critiques.
- The strategy memo is the contract for later execution.
- Catch design problems before code is written.
- Multiple candidate strategies are acceptable, but the approved path must be explicit.
- The user resolves strategic deadlocks after the third failed round.