Professional safety configuration by the cc-safe-setup team
In April 2026 alone:
Rolling out Claude Code across a team or organization? See the 法人・チーム向けサービス (Japanese) — org-wide policy, training, and CI safety gates.
August 2026 notice: none of the services below are being accepted
Every item here is work that a person has to take on per order, and the full package includes 30 days of ongoing support. That capacity isn't something I can responsibly commit to, so these are not accepted. If you came here to order one, I'm sorry.
What is available without any inquiry: the Team Safety Rollout Pack (¥3,000) on BOOTH, the free Security Checkup, and the hooks themselves (cc-safe-setup, MIT — free, always).
Review your settings.json, CLAUDE.md, and hook configuration. Identify vulnerabilities including deny rules bypass, PII leaks, and token waste.
Deliverable: Vulnerability report + recommended hooks + settings.json template
$50
Analyze your CLAUDE.md, context management, and workflow for token efficiency. Based on 800 hours of measured data.
Deliverable: Optimized CLAUDE.md + hook set + before/after comparison guide
$75
Build PreToolUse/PostToolUse hooks tailored to your specific workflow. Financial operations, data protection, CI/CD safety, or any custom requirement.
Deliverable: Tested hooks + settings.json integration + documentation
$100
Complete safety configuration: audit + token optimization + custom hooks + ongoing support for 30 days.
Deliverable: Everything above + 30-day email support
$200
cc-safe-setup author (914 example hooks) 800+ hours autonomous operation 170,000+ Qiita views on Claude Code
Corrected on 2026-08-09. Three certification claims that used to sit here — Anthropic Academy, Google Analytics, HubSpot Inbound — have been removed. Two of them were in fact obtained, but not in a way that supports listing them as a credential line: the Google Analytics (84%) and HubSpot Inbound (90%) exams are open-book by design, and both were taken with an AI advising on the answers. For the Anthropic Academy line I cannot reconstruct how it was obtained at all — there is a script here written to complete the coursework automatically, but no log that it ever ran, so I don't know what the claim rests on. Stating results without stating how they were reached invites the reader to assume something that may not have happened, so all three are out. The Qiita figure was also stale (80,000 → 170,000).
Free diagnostic first: Security Checkup (6 questions, 2 minutes).
For service inquiries:
Contact via GitHub