Spar — Development skill for Claude Code
Make the models argue before you build — Codex attacks the plan in bounded read-only rounds, then one builds and the other grades the diff.
How to install Spar
Installs to ~/.claude/commands/gavishap-cortex-spar.md
mkdir -p ~/.claude/commands && curl -fsSL https://raw.githubusercontent.com/gavishap/cortex/HEAD/.claude/commands/spar.md -o ~/.claude/commands/gavishap-cortex-spar.md Restart Claude Code, or start a new session, for it to be picked up.
What Spar does
description: Make the models argue before you build — Codex attacks the plan in bounded read-only rounds, then one builds and the other grades the diff argument-hint: [rounds=N]
Run the `sparring` skill on: $ARGUMENTS
The short version of what that means:
- SCOUT — recon from the vault's layered memory (relay status → graph → wiki catalog → narrow raw reads), then present ONE batched Assumptions Ledger with sources. Start state: `python scripts/spar_too
Alternatives in Development
- Fidelity — Measure how faithfully a clone reproduces a site — pixel-diff plus motion-fidelity into one 0-100 score, a let 3.6k ★
- Magnitude — Open source inference server your agent sets up for you 1.5k ★
- CI Mockup Figure — Read and follow the skill definition 239 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Adw Challenge
ADW step 3 - a fresh agent attacks the plan before any code is written
Cross Repo Authorize
Run the cross-repo authorization ceremony — restate the bounded action + target, get the user's yes/no, then w
Builder Self Review
Builder step 5 — re-read your own diff before publishing
The Docket
the-docket is a Claude Code plugin that keeps a repo's decisions from being undone. A hook prints what governs
Claude Code SDK TS
Configure models, enable tools, stream events, then fetch text, JSON, run details or token stats in one call v
Codex Review
Run a Codex Council-of-models review on the current diff and produce a triaged summary
Related Agents
Verifier Medium
Independently audits a diff against an approved PLAN.md — never against the implementer's own summary of what
Plan Challenger
Adversarial plan review before implementation. Attacks plans across 5 dimensions (Assumptions, Missing Cases,
Codex Cross Reviewer
Cross-review orchestrator using Codex CLI as 2nd reviewer. Runs structured finding exchange with consensus loo