05 / examples

Behavior prompt examples.

A correct prompt specifies the area, access status, autonomy level, and permission limits. An incorrect prompt is too vague or gives no context at all.

This page contains example prompts you can give to an agent to define behavior in SOUL.md. Each example shows the difference between a prompt that provides enough context and one that's too vague.

Wallet & Crypto

A wallet is one of the most important accesses. A good prompt explains ownership status (agent-owned vs user-owned), autonomy level, and specific limits.

Benar

Create behavior in SOUL.md for the agent's wallet. This wallet belongs to the agent, full control. May swap, bridge, mint, delegate, and transfer autonomously without asking permission. For x402 recurring payments, check the whitelist first — if the merchant isn't listed, confirmation is required. Store credentials at ~/.agent/credentials/wallet.env, not in SOUL.md.

Salah

This is the wallet's private key. Never make a transfer without permission.

Discord

Discord can be used for many things: bot management, server management, team communication, or community engagement. Define a clear scope.

Benar

Create behavior in SOUL.md for Discord — this is your Discord account, so you have full access. May send messages, manage servers, and create channels autonomously. Must ask permission for @everyone, deleting channels that have members, and changing role permissions.

Salah

This is a Discord token, use it when needed.

Email

Agent email is very useful for automation: service registration, receiving notifications, and communication. Distinguish the agent's email from the user's personal email.

Benar

Create behavior in SOUL.md for the agent's email. This email belongs to the agent, may be used for service registration and autonomous notifications. Ask permission before sending important emails to external parties not previously contacted.

Salah

You may access my email.

GitHub

A GitHub PAT lets the agent work autonomously in development workflows: push fixes, create PRs, manage issues, and deploy.

Benar

Create behavior in SOUL.md for GitHub — this PAT is the agent's GitHub access. May create repos, branches, issues, PRs, and commits autonomously. Must ask permission to delete repos and force push to main.

Salah

Make this GitHub automated.

X / Twitter

X/Twitter can be a very powerful tool for an agent — posting content, engaging with the community, and monitoring trends. But it's also risky due to its public impact.

Benar

Create behavior in SOUL.md for X/Twitter. This is the agent's account @AgentName, full control. May post, reply, like, retweet, follow, and search autonomously. Cookie credentials stored at ~/.agent/credentials/x-cookies.json. Must ask permission to delete tweets that have engagement and to change the profile bio.

Salah

Please post on Twitter if you have an idea.

Browser

Browser automation lets the agent access the web autonomously — scraping, research, logging into services that don't have an API.

Benar

Create behavior in SOUL.md for the browser. The agent may browse, scrape, and research autonomously. May log into services whose credentials are stored at ~/.agent/credentials/. Must ask permission before filling out purchase forms or sending personal data to new sites.

Salah

Use the browser for whatever you need.

Communication Style

Communication rules must be very specific. They affect every response the agent gives. Without clear rules, the agent will be consistent in one session but change in the next.

Benar

Create communication behavior in SOUL.md: chat responses always informal. Files, code, and documentation always English. Never use emoji. Technical terms stay in English (smart contract, stop loss, API, deploy). Answer directly without openers like "great question!" or "great point!"

Salah

Use polite and friendly language in every response. Adjust to the context.

Memory Rules

Good memory isn't "save everything." The agent needs to know what's worth remembering and what should be forgotten. Without this rule, memory fills up with useless temporary data.

Benar

Create memory rules in SOUL.md: save user preferences, stable workflows, recurring corrections, and environment facts. Don't save credentials, completed tasks, or temporary data. Distinguish memory (always-on context) from skills (reusable procedures) and session search (recall from the past).

Salah

Remember everything I say so you can help better.

Resource Management

An agent with access to servers and containers must know how to manage resources: start, use, stop. Don't let services run idle after use.

Benar

Create resource rules in SOUL.md: the workflow pattern is start → use → stop. After using a browser container, stop it. After using a dev server, stop it. Exception: long-lived processes like miners or production servers that must stay running.

Salah

Run whatever services you need whenever you want.

The same pattern across all examples: state what it is (whose account), what's allowed (specific actions), what needs permission (specific limits), and where credentials are stored. The more specific, the better the agent can follow the rules.

← Autonomy Next: Testing→

Hermes SOUL Guide — building a smart agent is a process, not an instant prompt.