Codex Strategy Review

Independent adversarial logic review using Codex (gpt-5.6-sol, model_reasoning_effort=xhigh). This skill is generic by default and flexible for different personas.

Requires the codex CLI (npm install -g @openai/codex, then codex login). Persona listing/resolution works without it.

Focus

Use this to stress-test a strategy, argument, proposal, research conclusion, risk framework, optimizer design, decision memo, or any other plan where the question is "does this logic hold up?"

The generic review checks:

• conclusions that do not follow from premises
• hidden, circular, fragile, or unstated assumptions
• made-up thresholds, constants, weights, or scoring rules
• missing evidence, counterexamples, and falsification tests
• incorrect math, units, base rates, or time horizons
• feasibility, incentives, adoption, and operational constraints
• downside, tail risks, failure modes, monitoring, and reversibility
• contradictions between goals, constraints, and proposed mechanisms

Personas

Personas are optional lenses, not a fixed enum. Decide which to use up front when the review is initiated. Example lenses: architect, cto, red-team, quant, portfolio-manager — or pass a direct path to a persona .md file.

Run without personas for a generic review; pass --personas a,b,/path/to/persona.md for a panel.

List what's available via the shared resolver (works even without Codex):

python3 ~/.claude/skills/persona-registry/scripts/persona_registry.py --list

Model & reasoning effort

Defaults: model gpt-5.6-sol, reasoning effort xhigh. To change (no reinstall): pass --model / --effort, or edit DEFAULT_MODEL / DEFAULT_EFFORT at the top of the script.

Workflow

Generic review:

python ~/.claude/skills/codex-strategy-review/scripts/codex_strategy_review.py \
  <strategy_file> \
  [context_file ...]

Generic review plus persona overlays:

python ~/.claude/skills/codex-strategy-review/scripts/codex_strategy_review.py \
  <strategy_file> \
  [context_file ...] \
  --personas quant,pm,/path/to/custom_persona.md

Legacy form still works:

python ~/.claude/skills/codex-strategy-review/scripts/codex_strategy_review.py \
  <strategy_file> \
  quant,pm \
  [context_file ...]

Each persona pass is independent. The final synthesis may see all completed outputs and deduplicates agreement, disagreement, and unique catches.

Output Contract

Generic:

## Codex Strategy Review — Generic Logic

Persona:

## Codex Strategy Review — <Persona Name>

Synthesis:

## Codex Strategy Review — Panel Synthesis

Persona resolution

Personas resolve via the shared resolver (persona-registry/scripts/persona_registry.py), which is installed regardless of Codex. The Codex script locates it automatically. Search order (first existing wins):

1. $PERSONA_REGISTRY_DIR (explicit override, if set)
2. co-located with the Codex script
3. the skill dir / skills root (repo layouts)
4. the sibling persona-registry skill (in-repo layout)
5. ~/.claude/skills/persona-registry/scripts (installed canonical resolver)

Persona discovery (which .md files are available) is governed by the persona-registry skill: PERSONA_PATHS / --persona-dir, then the installed kit library (~/.config/persona-review-kit/personas/), then project-local folders, then ~/.claude/personas/.

Intent and focus (automatic)

Derive these from the user's request — they should not need to type flags. From how they phrase the ask, infer intent (what the work is trying to do) and focus (what to weight) and pass them as --intent / --focus. For example, "review the auth refactor on this branch, focus on whether the logic is right and skip style" -> --intent "auth refactor" --focus "logic correctness, skip style".

The text is appended to the prompt as author framing, never as another reviewer's findings (the review stays independent). A guardrail still surfaces every CRITICAL/HIGH finding even when it falls outside the focus.