jwforge

One-time project analysis — writes a split set of files under .jwforge/ (overview + module map, plus optional api-map / data-model). Persistent across /work cycles.

Brainstorm what to build — writes .jwforge/current/brainstorm.md

Build a TDD-ready plan from brainstorm (+ optional analysis). Annotates each item with the files it will touch and assigns a wave number for parallelizable execution. Iterates with user until approved.

Re-run the lightweight review (security + lint + light scan). Useful after manual edits.

Execute the approved plan wave-by-wave through the TDD pipeline. Runs items via tdd-item-runner. Optional parallel execution within a wave (gated by pre-flight check). Followed by lightweight review.

Analyzes an existing codebase to produce a split set of analysis files under .jwforge/. Has two modes — full (first time) and incremental (called by /work after changes). Used by /jwanalyze and /work.

Writes the minimum production code required to make a given set of failing tests pass (the Green step of TDD). Never modifies test files.

Runs the full ①→⑤ TDD pipeline for a single plan item inside a given working directory. Used by /work for both single-item and multi-item waves. Returns a structured summary including the actual files modified.

Refactors production code written in the Green step while keeping all tests passing. Never touches test files or the spec. Returns a diff summary.

Produces a focused list of test cases (normal, boundary, failure) for a TDD slice, and independently verifies coverage after green. Used in stages ① (spec) and ⑤ (verify) of the jwforge TDD pipeline. Do not use for implementation.