@peruntechllc/ai-engineering

PerunTeCH AI Engineering

The audit trail your AI coding agents don't leave behind.

@peruntechllc/ai-engineering installs SDLC governance into AI dev environments. It is not an AI assistant — it is the discipline layer on top of the codegen tools you already use. Built for Claude Code today; the pre-commit gate sits at git level, so it holds no matter what wrote the code — Cursor, Codex, or a human. Inside: a pre-commit hook that blocks any commit lacking a change-request (CR) ticket reference, post-write scans that flag secrets, OTPs, and category-sensitive data in AI-written files (Claude Code hooks), a session-start hook that surfaces the open review backlog for seat-to-seat handoff, plus CR-tracker spreadsheet templates — the who-what-which-ticket-verified-how trail a SOC 2 auditor asks for — a memory-file taxonomy, and review-gate playbooks (CCPA/CPRA, COPPA, DPDP — engineering checklists, not legal advice). The pack runs entirely against your repo, and every hook that executes ships as readable source in this package — audit it before you enable anything. Only a license key, a machine-fingerprint hash, and the pack version are ever transmitted; never your code. Hooks require an active subscription key (see LICENSE.md); without one they no-op with a visible message and never break your commit flow.

This is not a course, not a consulting subscription. It is software. Subscribe once; the practice becomes a forcing function in your AI workflow from then on.

Pricing

USD 49 per seat per month. Commercial subscription license. The pack's hooks and skills require a valid license key to function — no free tier. Cancel anytime; non-refundable after 14 days from initial purchase.

→ Subscribe + get a license key

14-day money-back guarantee. Try the full pack risk-free; if it doesn't ship cleaner SDLC discipline through your AI workflow in two weeks, request a refund through the billing portal and we cancel + refund the first month, no questions.

Why this exists

AI dev tools default to vibe coding — generate code, iterate, ship. That works for prototypes and falls apart on production systems. Vibe coding skips SDLC.

PerunTeCH AI Engineering makes that hard to do. The pack ships:

• Skills that prompt the model to capture a requirement before generating code, write an SH-Review handoff before yielding the seat, and update the CR tracker on every commit.
• Hooks that block commits without a ticket reference, run a security redaction sweep on every file write, and surface seat-rotation context on session start.
• Memory templates so context survives across sessions / agents / seats — the way it does in a real engineering org.
• CR-tracker templates — the Excel schema PerunTeCH itself runs on (Defects, Enhancements, Dashboard, status taxonomy: Open → In Progress → ENGR Review → SH Review → Closed).
• Review-gate playbooks for security sweeps, redaction patterns, compliance posture (DPDP, CCPA / CPRA, COPPA), and pre-launch checks.

Reference implementation

The six-surface PerunTeCH studio stack — sattanaathan.com, peruntech.ai, peruntechsystems.com, ranklens.in / ranklens.online, peruntechauto.in / peruntechauto.online — is the reference implementation. Every gate, every template, every hook in this pack has shipped production surfaces against the SDLC described in INSTALL.md. The studio dogfoods this pack; that is the warranty.

What's in the box

@peruntechllc/ai-engineering/
├── README.md                       ← this file
├── INSTALL.md                      ← step-by-step install + license-key setup
├── LICENSE.md                      ← Commercial Subscription License
├── package.json                    ← npm package metadata
├── settings.json.template          ← Claude Code settings + hook wiring
├── skills/
│   ├── cr-tracker-update/SKILL.md  ← updates the CR tracker on every commit
│   ├── sh-review-handoff/SKILL.md  ← writes the SH-Review handoff note
│   ├── security-sweep/SKILL.md     ← human-shape security review before SH-Review / merge
│   └── llm-review/SKILL.md         ← optional advisory 2nd-LLM smell-test (bring-your-own endpoint)
├── reviewer/
│   ├── llm_review.py               ← the 2nd-LLM reviewer (model-agnostic; no-ops without an endpoint)
│   └── test_llm_review.py          ← pytest suite for the reviewer
├── hooks/
│   ├── pre-commit-cr-trace.sh      ← blocks commits without a ticket reference
│   ├── parse_commit_msg.py         ← helper: extracts command + commit message from CLAUDE_TOOL_INPUT
│   ├── post-write-redaction.sh     ← PostToolUse: redaction sweep on Write/Edit
│   ├── redaction_scan.py           ← helper: secret / Auth0 sub / OTP / category-field scanner
│   ├── session-start-cr-context.py ← SessionStart: surface SH-Review backlog as primer
│   └── test_redaction_scan.py      ← pytest suite for the redaction scanner
├── memory-templates/
│   ├── README.md                   ← the taxonomy (project_*, feedback_*, reference_*)
│   ├── CLAUDE.md.template          ← root project-instructions file (loaded into every session)
│   ├── project_overview.md.template
│   ├── feedback_template.md.template
│   └── reference_template.md.template
├── cr-tracker-templates/
│   ├── README.md                   ← schema, status taxonomy, conventions
│   └── tracker-blank.xlsx          ← Defects + Enhancements + Dashboard sheets
└── examples/
    └── reference-deployment-pointer.md

Install (after subscribing)

npm install --save-dev @peruntechllc/ai-engineering
export PERUNTECH_AI_ENGINEERING_LICENSE_KEY=<the key from your subscription email>

Then follow INSTALL.md to wire the hooks, skills, and CR tracker into your project. The hooks validate the key against PerunTeCH's license endpoint — only the key, an anonymous machine fingerprint hash, and the pack version are transmitted; nothing about your code or repository. Results are cached for 24 hours, and a 7-day offline grace window keeps the pack working without connectivity. When the subscription ends or the key can't validate past grace, the hooks visibly no-op with a billing-portal notice — your commits are never broken.

Want a configured install or a studio engagement?

The $49/mo subscription gives you the standard pack. If you'd rather have it tuned to your stack, AI tool, compliance posture, and team workflow — and the studio standing behind the configuration — that's a paid engineering engagement, billed separately.

→ peruntech.ai/contact?offer=ai-engineering

Who ships this

Operated by PERUNTECH LLC (US LLC, US/global commerce). The India sister entity PERUNTECH SYSTEMS (OPC) PVT LTD. (peruntechsystems.com) is the studio surface where the reference implementation lives. Same product across both entities; ships from whichever entity is closer to the buyer's jurisdiction.

License

Commercial Subscription License. See LICENSE.md. Paid subscription + valid license key required for use; no free tier; no redistribution. Refundable within 14 days.