rajphanee

Enterprise Agent Platform — Security skill for Claude Code

Security community

Three governed enterprise AI agents (permission-aware RAG, IT support, vendor onboarding) with a code-based policy engine, human approvals, and a hash-chained audit log.

How to install Enterprise Agent Platform

This entry records only its repository, not the path inside it, so there is no exact command to give. Open rajphanee/enterprise-agent-platform and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Enterprise Agent Platform does

Three governed enterprise AI agents (permission-aware RAG, IT support, vendor onboarding) with a code-based policy engine, human approvals, and a hash-chained audit log. Runs offline or on Claude.

Alternatives in Security

OpenTag — Open-source, channel-native agent gateway for Slack 499 ★
Senior Engineering Partner — A stack-agnostic Claude Code skill: strict code reviewer, pair programmer, debugger, and mentor (Python/Bash/A 148 ★
CI Audit — Stack-aware CI gap detection — audits a project's GitHub Actions workflows against its declared tech stack and 117 ★

README

Enterprise Agent Platform

A small, readable reference implementation of three enterprise AI agents that share one governed platform:

Knowledge assistant – permission-aware Q&A. Every sentence must cite a source document; if it can't, it declines instead of guessing.
IT support agent – an LLM classifies the ticket, but a policy engine written in code decides what runs automatically, what waits for a human approval, and what goes straight to an engineer.
Vendor onboarding agent – deterministic checks (tax ID, bank name match, routing checksum, insurance, duplicates, watchlist), parallel finance/legal review, and an LLM-drafted fix list for the vendor.

Every decision from every agent is written to a **hash-chained audit log** that detects edits and deletions.

It runs **fully offline** (no API key, no model download, only `pytest` to install) for development, tests, and CI. Set `AGENT_MODE=claude` to use the Anthropic API instead.

1. Folder layout

enterprise-agent-platform/
├── agent_platform/            # the application package
│   ├── app.py                 # build_platform(): wires everything together  <- START HERE
│   ├── cli.py                 # command line (python -m agent_platform ...)
│   ├── config.py              # all settings, read from env / .env
│   ├── errors.py              # every expected error type
│   ├── audit.py               # hash-chained audit log
│   ├── llm.py                 # OfflineLLM + ClaudeLLM (the only file that talks to a model)
│   ├── knowledge/
│   │   ├── store.py           # TF-IDF search with role filter applied BEFORE scoring
│   │   └── assistant.py       # citation gate, per-role cache, declines
│   ├── it_agent/
│   │   ├── policy.py          # routing rules (code, not prompt)
│   │   ├── approvals.py       # approval queue + separation of duties
│   │   └── agent.py           # ticket flow + simulated automations
│   └── vendor/
│       ├── validation.py      # deterministic check