Skip to content
Back to search

greenfield

SkillWorks with Claude CodeGitHub

A Claude Code plugin that reverse-engineers clean behavioral specs, test vectors, and acceptance criteria from any codebase, producing a provenance trail so a fresh team can reimplement without inheriting the original's internal structure.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/prime-radiant-inc-greenfield/badge.svg)](/listings/prime-radiant-inc-greenfield)

HTML

<a href="/listings/prime-radiant-inc-greenfield"><img src="/api/listings/prime-radiant-inc-greenfield/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

/plugin marketplace add prime-radiant-inc/prime-radiant-marketplace
/plugin install greenfield@prime-radiant-marketplace

Restart Claude Code after installing.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 80 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Greenfield
  2. **Reverse engineer clean behavioral specs from any codebase.**
  3. Greenfield reads source code, documentation, SDKs, runtime behavior, and binaries, then produces behavioral specifications, test vectors, acceptance criteria, and a full provenance trail. The output describes *what* the software does — not *how* any particular codebase does it — so a fresh implementation team can build against the specs without inheriting the original's internal structure.
  4. Greenfield is a [Claude Code](https://claude.com/claude-code) plugin. It runs inside the `claude` CLI and uses Claude agents as the workers that read code, write specs, and audit output.
  5. ## Installation
  6. ```bash
  7. /plugin marketplace add prime-radiant-inc/prime-radiant-marketplace
  8. /plugin install greenfield@prime-radiant-marketplace
  9. ```
  10. Restart Claude Code after installing.
  11. ## Usage
  12. ### Extract specs
  13. ```bash
  14. claude
  15. > /analyze /path/to/target
  16. ```
  17. `/analyze` runs a seven-layer pipeline. It discovers the available intelligence sources (source, docs, SDK, community, runtime, binary, git history, tests, UI, contracts), gathers evidence from each, synthesizes behavioral specs with provenance citations, generates test vectors and acceptance criteria, sanitizes the specs of implementation details, then runs a second-pass review of the result.
  18. Target shape doesn't matter. Greenfield has been used on single-file minified JavaScript bundles and on source-tree projects; the pipeline also has a path for decompiled native binaries. The methodology adapts to what's there.
  19. Workspace output:
  20. ```
  21. workspace/
  22. ├── raw/ # Analysis artifacts with source references
  23. ├── output/ # Sanitized specs for the implementation team
  24. │ ├── specs/
  25. │ ├── test-vectors/
  26. │ └── validation/
  27. └── provenance/ # Citation audit trail
  28. ```
  29. ### Re-sanitize an existing workspace
  30. ```bash
  31. claude
  32. > /sanitize /path/to/workspace
  33. ```
  34. Re-runs the sanitization pass on an existing workspace. Useful when the initial pass left contamination that the audit caught.
  35. ### Implement from the output specs
  36. Greenfield stops at the specs. The implementation team reads `workspace/output/` and builds against the behavioral specs, test vectors, and acceptance criteria there.
  37. ## How it works
  38. The pipeline dispatches agents with role-specific prompts across the seven layers. Each dispatch runs one of two generic agent types loaded with a specific skill:
  39. - **analyzer** — the workhorse. Reads evidence and writes specs. Dispatched under many roles across layers: `discovery-agent`, `bundle-splitter`, `chunk-analyzer`, `function-analyzer`, `doc-researcher`, `community-analyst`, `sdk-analyzer`, `cli-explorer`, `behavior-observer`, `feature-discoverer`, `architecture-analyst`, `api-extractor`, `synthesizer`, `module-mapper`, `deep-dive-analyzer`, `behavior-documenter`, `user-journey-analyzer`, `contract-extractor`, `spec-verifier`, `source-completeness-checker`, `test-vector-generator`, `test-generator`, `acceptance-criteria-writer`, `spec-reviewer`, `structural-leakage-reviewer`, `content-contamination-reviewer`, `behavioral-completeness-reviewer`, `deep-read-auditor`, `fidelity-validator`.
  40. - **sanitizer** — rewrites raw specs into output specs free of implementation details. Dispatched during Layer 5 and for remediation during Layers 6 and 7.
  41. ### Pipeline
  42. | Layer | Roles | Output |
  43. |-------|-------|--------|
  44. | L1: Intelligence | bundle-splitter, chunk-analyzer, function-analyzer, doc-researcher, community-analyst, sdk-analyzer, cli-explorer, behavior-observer | Raw evidence from source, docs, SDK, community, runtime, binary |
  45. | L2: Synthesis | feature-discoverer, architecture-analyst, api-extractor, synthesizer, module-mapper | Feature inventory, architecture model, module map |
  46. | L3: Deep Docs | deep-dive-analyzer, behavior-documenter, user-journey-analyzer, contract-extractor | Behavioral specs, journeys, contracts |
  47. | Gate 1 | spec-verifier | Correctness, contradictions, gaps |
  48. | Gate 1b | source-completeness-checker | Every user-facing surface captured |
  49. | L4: Validation | test-vector-generator, test-generator, acceptance-criteria-writer | Test vectors, test specs, acceptance criteria |
  50. | Gate 2 | spec-reviewer | Implementation leakage, completeness, quality |
  51. | L5: Sanitization | sanitizer | Output specs rewritten from understanding, not copied |
  52. | L6: Review | structural, content, completeness reviewers + deep-read auditors | Second-pass contamination review |
  53. | L7: Fidelity | fidelity-validators | Flags behavioral detail lost or weakened during sanitization |
  54. ## License
  55. Apache 2.0. Copyright 2026 Prime Radiant, Inc.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.