AVAILABLEAI Agent Template Pack
Available product — files, demo URL and secure download get mapped in EDF 0.3.
What it does
Twelve agents you can actually ship.
Twelve AI agent templates that are more than a prompt: customer support triage, refunds with approval, lead qualification, meeting notes and actions, cited research, pull request review, read-only SQL analysis, editorial assistance, customer onboarding, invoice extraction, incident first response and fair candidate screening. Each template is a complete, reviewed definition: the system prompt with variables (role, procedure, rules, when to hand over, output format), tools with strict argument schemas, a risk level and an approval rule, guardrails, a memory policy and eval cases (normal, edge and refusal). Around them sits a working runner that enforces what a prompt cannot: schema validation of every tool call, human approval before risky actions, step and call limits, loop stop and optional guard hooks, plus adapters for OpenAI-compatible APIs and Anthropic, a client for AI Guardian, a quality gate, a CLI and a browser Studio where you inspect a template, edit its variables, replay its eval cases and approve or decline the agent's actions yourself. Node.js 22+, no npm dependencies, 43 automated tests. The eval replays use a scripted model, so they prove the template and the runner are consistent, not how your model behaves; the guide shows how to run the same cases against your own model. Mock tools use fictional data.
Origin: EDF native product (Platform Modules 1.0) · Roadmap window:
Live demo


Package
Placed in the private vault layout (deploy/vault/products). Downloads stay disabled until checkout is connected.
What's included
- 12 templates across support, sales, operations, research, engineering, data, content, customer success, finance and HR, each with variables, tools, approval rules, guardrails, hand-over conditions, memory policy and 4-5 eval cases (49 in total)
- Runner: agent loop with schema validation of every tool call, human approval for risky actions (a decline is reported back to the model), step and tool-call caps the caller cannot raise, loop detection, optional guard hooks on input, tool calls and output
- Quality gate (lint): refuses high-risk tools without approval, schemas that allow extra arguments, prompts without rules or a hand-over section, missing mocks, too few eval cases, no refusal case
- Adapters for OpenAI-compatible APIs (OpenAI, Azure-style gateways, OpenRouter, vLLM, Ollama, ModelRouter) and Anthropic, tested against scripted HTTP responses; a client for AI Guardian that waits for approvals
- Exports: OpenAI and Anthropic tool definitions and request bodies, a ready AI Guardian tool registry (same risk and approval rules), a one-page Markdown description
- Studio in the browser: gallery with search, prompt with live variables, tools and schemas, guardrails, a run console where you approve or decline the agent's actions and see the scripted agent take the other branch, export
- CLI: list, show, prompt, export, validate, eval, and run against a real model with terminal approvals
- 43 automated tests, full source and docs: testing with your own model, writing templates, safety notes
FORGE DNA™
TIMEPROOF SCORE — derived from real catalog fields (stack, formats, skill, AI flag), not decoration
Compatibility & relations
Derived by the mapping engine (family · layer · tags · dependency rules). Stored in the catalog as compatibility, dependencies, relations[].
Documentation
- Quick start (planned — every product gets one)
- Customisation guide
- Deployment guide — Vercel / Cloudflare / GitHub
- epic.json / EPIC-DNA.json — machine-readable DNA shipped inside the ZIP
Version history & changelog
| VERSION | STATE | NOTES |
|---|---|---|
| v1.0.0 2026-10-13 | AVAILABLE | Initial release. This package contains: 12 templates across support, sales, operations, research, engineering, data, content, customer success, finance and HR, each with variables, tools, approval rules, guardrails, hand-over conditions, memory policy and 4-5 eval cases (49 in total); Runner: agent loop with schema validation of every tool call, human approval for risky actions (a decline is reported back to the model), step and tool-call caps the caller cannot raise, loop detection, optional guard hooks on input, tool calls and output; Quality gate (lint): refuses high-risk tools without approval, schemas that allow extra arguments, prompts without rules or a hand-over section, missing mocks, too few eval cases, no refusal case; Adapters for OpenAI-compatible APIs (OpenAI, Azure-style gateways, OpenRouter, vLLM, Ollama, ModelRouter) and Anthropic, tested against scripted HTTP responses; a client for AI Guardian that waits for approvals; Exports: OpenAI and Anthropic tool definitions and request bodies, a ready AI Guardian tool registry (same risk and approval rules), a one-page Markdown description; Studio in the browser: gallery with search, prompt with live variables, tools and schemas, guardrails, a run console where you approve or decline the agent's actions and see the scripted agent take the other branch, export. |
| → | Roadmap | Target window for the next stage. |
License
Standard (personal or commercial use in one end project or your own business) · Extended ×3 (SaaS, unlimited client projects, finished products only). Agency licence comes later.
Deployment
Activated by 1.0 (download) and 2.0 (one-click deploy). Download will be a 10-minute signed URL from the R2 vault — never a public /files/product.zip.