Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add bengous/claude-code-plugins --skill greenfield-auditgit clone --depth 1 https://github.com/bengous/claude-code-pluginsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/bengous/claude-code-plugins/greenfield-audit)<a href="https://agentmods.dev/skills/bengous/claude-code-plugins/greenfield-audit"><img src="https://agentmods.dev/badge/skills/bengous/claude-code-plugins/greenfield-audit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bengous/claude-code-plugins/greenfield-audit"><img src="https://agentmods.dev/badge/skills/bengous/claude-code-plugins/greenfield-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00054 | $0.01708 |
| Opus 5.5 | $0.00022 | $0.00683 |
| Sonnet 5.5 | $0.00011 | $0.00342 |
| Haiku 4.5 | $0.00005 | $0.00171 |
Grade A, and why
greenfield-audit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Greenfield audit
Scope: $ARGUMENTS
When the scope is empty, ask for the path in the chat and stop. This is the only question the run may ask.
Stance
The current architecture has no authority. Several generations of agents produced and modified it. Expect conceptual debt, accidental abstractions, and inherited decisions whose justification no longer exists.
The goal is not to make the current code more elegant. The goal is to find the minimal architecture the system needs today and to measure the existing code against it.
Constraints for the whole run:
- Read-only. The editing tools are removed for the run; do not write through Bash either. The diagnostic goes in the chat.
- Do not ask questions mid-run. Collect them in the final "Unknowns" section.
- Cite evidence as
file:line, or as a commit hash for intent recovered from git, for every requirement, invariant, and verdict. - Follow the phases in order. Emit the Phase 1 lists and the Phase 2 design as visible text before the first Phase 3 tool call. Code that has been read cannot be unread, so this order is the only protection against anchoring on the existing design. A Phase 2 design that appears after an implementation file was read is a violation.
Copy this checklist into the response and tick each line when its text is emitted:
Audit progress:
- [ ] Phase 1: five requirement lists and Unproven emitted, no implementation file opened
- [ ] Phase 2: target architecture emitted, still no implementation file opened
- [ ] Phase 3: implementation read, important components inventoried
- [ ] Phase 4: one verdict per component with evidence
- [ ] Output: five sections delivered, no file edited
Phase 1: Requirements
Establish what the system must do from sources that describe behavior, not from the code that implements it.
Allowed sources in this phase:
- Tests, fixtures, snapshots
- Public entry points: routes, CLI commands, exported API, UI screens. Open an entry point only to list what it exposes (route table, command names, exported signatures). Stop at the signature; do not follow its imports.
- Persisted schemas, migrations, seed data
- Configuration, environment variables, deployment files
- README, docs, ADRs, issues referenced in the tree
git log --onelineor--staton the scope, and commit messages, to recover intent. No-p, nogit show, nogit blame: they print implementation.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 174 lines · 54 tokens per session scan A cb34facb0ac8
greenfield-audit is a skill published in the GitHub repository bengous/claude-code-plugins (5 stars, last pushed today), licensed MIT. It adds 54 tokens to every session and 1,708 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-11.
Other skills, from other repositories
verify-change
A change-checking skill that reviews code differences and checks whether related documentation, tests, and records were updated.
verify-module
A module-checking skill that scans a project’s folders, code, and documentation. It expects each module to include a README.md and DESIGN.md, which explain how the module is used and why it was designed that way.
verify-quality
A code-quality checker that measures complexity, size, naming, and common code smells. It reports problems such as duplicated code, unused code, and overly long functions.
planning
ONLY for coordinated multi-artifact work: multiple workflows with dependencies, shared data-table schema/migration across tasks, or the user explicitly asked to review a plan first. If create-tasks is not visible, load it via loadtool (search "create tasks"). Do NOT use for new one-off workflows, single-workflow…
criticism-self-criticism
A structured review method for examining completed work, criticism, and repeated mistakes. It focuses on specific evidence, causes, effects, and practical improvements.
flow-next-impl-review
Carmack-level implementation review of changes via the configured backend. Use when asked to review code or a diff in a flow-next repo.