home/productivity/copilot-instructions-182

Copilot Instructions

GPTClaudeDeepSeek··▲ 852 copies·updated 2026-07-14

copilot-instructions-182.prompt

# Copilot instructions — agentic-handbook

This repository is a vendor-neutral, CPTO-grade playbook for shipping reliable agents in production.
The seven surfaces of a reliable agent system spell **AGENTIC**: Autonomy, Goals, Evaluation, Networks,
Trust, Identity, Cost. When suggesting edits, keep the repo's hard rules.

## Hard rules

1. **Synthetic / safe only** — fictional orgs, toy datasets, no real customer data.
2. **Label volatile facts** — every star count, benchmark %, CVE, dollar figure, or date carries an
   `as of June 2026 — verify before relying` label.
3. **Cite external claims** — each one gets a row in [`../references.md`](../references.md); citations
   link `references.md#<id>` and must not dangle.
4. **One canonical mapping** — the AGENTIC↔chapter map and the capability-ladder↔EU-AI-Act risk-class
   mapping live once in [`../repo.config.json`](../repo.config.json); never retype them out of sync.
5. **Receipts match prose** — example headline numbers are machine-checked against committed receipts.

## How to verify a change

Run the unit validators and the render-check before proposing a change is done:

when to use it

Community prompt sourced from the open-source GitHub repo robinjose911/agentic-handbook (MIT). A "Copilot Instructions" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.