Agent Kit sets up your instructions and guard hooks. Repo Audit checks those instructions against your code.
For developers using Claude Code, Codex, Cursor or GitHub Copilot.
Choose Kit + Audit · €169Kit €29. Audit €149. One-time, no VAT charged (§ 19 UStG)
Read the real audit first →
In our audit of block/goose on 21 October 2026, 59 of 65 instruction claims matched the code; 6 needed a rewrite, and the repository scored 12.5 of 19 checkable points.
Inspect the nine-page report ↗Every instruction claim needs a source. Every correction needs a reason.
One public repository. Up to 10 instruction files / 1,200 instruction lines. For larger repositories, we email you before starting with what we will cover. Static review: we never run your code.
Score, top 3 fixes and how agents will experience the repo.
Paths, commands, scripts, packages and tool rules checked against the tree. Keep, rewrite with proposed text, or delete.
Tool-specific discovery and conflicting instructions, with a source of truth to keep.
Lines and bytes per file; what to cut or move into scoped guidance.
Evidence for each point. Repository settings that cannot be checked are marked separately.
Concrete file references, proposed changes and S/M/L effort estimates.
The goose desktop guide used the watch-mode test command. The audit points to the one-shot script in ui/desktop/package.json:36–37.
AGENTS.md:73
pnpm test
→ cd ui/desktop && pnpm run test:runThe audit reads your code and configuration. It never runs them.
A shared source of guidance. Separate controls where the tools support them.
.claude/hooks/agent-guard.sh
.claude/settings.json
.cursor/hooks.json
.cursor/mcp.json
.cursor/rules/00-core.mdc
.cursor/rules/10-edit-loop.mdc
.cursor/rules/20-testing.mdc
.cursor/rules/30-security.mdc
.cursor/rules/40-git.mdc
.cursor/rules/50-frontend.mdc
.cursor/rules/60-go.mdc
.cursor/rules/70-typescript.mdc
.github/copilot-instructions.md
AGENTS.md
CHANGELOG.md
CLAUDE.md
DOCS-CHECKED.md
README.md
WORKFLOW.md
checklists/agent-readiness.md
checklists/pr-review.md
prompts/01-plan.md
prompts/02-bugfix.md
prompts/03-refactor.md
prompts/04-review.md
prompts/05-test-backfill.md
prompts/06-handoff.md
scripts/agent-readiness.sh
tests/fixtures/paths.md
tests/run.shAgent-readiness scan: ~/src/goose
[x] AGENTS.md present
[x] AGENTS.md is concise (132 lines)
[x] AGENTS.md mentions test commands
[x] AGENTS.md states boundaries
…
[ ] glob-scoped rules -> scope rules with globs (Cursor only)
[ ] guard hooks -> copy .claude/hooks + .cursor/hooks.json from the kit
…
Score: 11/14 (heuristic scan; full 20‑point audit: checklists/agent-readiness.md)Excerpt from the Kit scanner run on block/goose on 21 October 2026. The scan has 14 automated checks; the full checklist in the Kit has 20 points.
Adapt the commands and paths, then check the installed permissions and hooks. Claude’s supplied failure-blocking setting needs v2.1.295+ and the guard needs jq.
| Layer | What it does | Where it stops |
|---|---|---|
| Instruction files | Guide the model. | Loading varies. Markdown is not an access boundary. |
| Claude permission rules | Deny matching tools and inputs. | Indirect script access needs a sandbox. |
| Claude / Cursor guard hook | Denies detected secret-looking paths; asks about detected destructive shell commands. | Configured events and text matching only. No called-script inspection, symlink resolution or MCP coverage. |
| Codex sandbox and approvals | Enforce execution access and gate escalation. | Configured separately. Workspace access does not hide workspace secrets. The Kit does not configure Codex hooks. |
| Required CI checks | Block merging when required by the repository. | The Kit installs no CI workflow. |
The Kit covers instructions and guard rails. For interface design and copy, Sans Slop adds UI and writing rules for the same agents, with a checker you can run in CI.
Use one if it fits. The work is connecting it to your repository.
The Kit combines the shared file, tool-specific setup, guard hook, six workflow prompts and checks. You replace the example commands and paths.
The Audit checks the actual tree, resolves conflicting guidance and writes concrete corrections. Read the sample to judge that work.
Public problem reports, not product reviews or endorsements.
“My CLAUDE.md specifies running analyzers before completion. Agents appear to ‘reason about’ code instead of actually executing the tools.”
joemc3 on GitHub, 15 January 2026 ↗
“This way, projects don't need to keep multiple duplicate text files for generative AI agents.”
brylie on GitHub, 13 August 2025 ↗
All prices in EUR. No VAT is charged under § 19 UStG (German small business regulation). One-time payment.
Set up your own repository.
€29
€39 from 1 November 2026
Kit download page right after payment, plus a confirmation email.
Buy Agent Kit · €29Get a checked worklist for one public repository.
€149
Report by email as PDF + Markdown within 48 hours of payment.
Buy Repo Audit · €149Recommended: reusable files and a review of your repo.
€169
Kit download page right after payment, plus a confirmation email. Report by email as PDF + Markdown within 48 hours of payment.
Buy Kit + Audit · €169Align instructions across three public repositories.
€399
Kit download page right after payment, plus a confirmation email. Audits by email as PDF + Markdown, plus the summary PDF, within 5 working days.
Buy Team Pack · €399One public repository. Up to 10 instruction files / 1,200 instruction lines. For larger repositories, we email you before starting with what we will cover. Team Pack applies that scope to each repository.
For an audit, enter the GitHub, GitLab or Codeberg URL at checkout.
Kit and bundle purchases open the download page right after payment. A confirmation email follows.
Single-repository audits arrive by email as PDF + Markdown within 48 hours of payment. The Team Pack’s three audits and cross-repo summary PDF arrive within 5 working days.
Email us before buying.
Claude Code, Codex, Cursor and GitHub Copilot, plus other agents through AGENTS.md. Loading and scope vary by tool. The Kit includes setup notes for each named tool.
No. The audit is a static review of instruction files, source, manifests and CI configuration. We never run your code.
The audit drafts an AGENTS.md from the code and repository configuration.
Yes. Enter your business details at checkout and Stripe emails the invoice. No VAT is charged under § 19 UStG (German small business regulation).
Free templates are a useful starting point. The Kit combines shared instructions, tool-specific setup, a guard hook, workflows, a read-only scan and regression tests. You adapt the commands and paths. The Audit checks your actual repository and proposes specific corrections.
Thank you. Your audit starts with the repository you entered. The report arrives by email within 48 hours.
Contact us