Self-Sufficient Agent

A general-purpose autonomous AI agent that combines shell execution, persistent space storage, and self-modification to handle virtually any task - adapting its own instructions over time and building on prior sessions without ever losing context.

←Back to examplesCopy
autonomous
self-sufficient
self-modifying
1464

The Self-Sufficient Agent blueprint represents the most capable single-agent configuration available on the ChatBotKit platform. It combines execution power, persistent memory, and self-modification into one cohesive agent that can genuinely handle any task thrown at it - from coding and research to automation and planning.

Most AI agents are constrained by three fundamental limitations: they cannot execute real actions, they forget everything between sessions, and their behavior is fixed at deploy time. This blueprint eliminates all three.

Execution via Shell

Three tightly integrated shell abilities give the agent real-world reach:

  • bash - Runs shell commands and scripts in an isolated sandbox. The agent can compile code, run tests, install packages, call CLIs, and surface the output directly in chat. Nothing is theoretical.

  • rw - Reads and writes file content in the shell environment with optional line-range targeting for surgical edits. The agent creates source files, patches specific sections of existing code, and reads back results for review.

  • import - Fetches content from HTTP/HTTPS URLs and saves it locally. Remote assets, starter templates, and external data files become part of the agent's working environment on demand.

Persistent Memory via Space Storage

The shell sandbox is ephemeral - files written there vanish when the session ends. The persistent space solves this. The agent is explicitly instructed to treat the space as long-term memory: reading prior state at the start of every relevant task, writing results and reusable assets back after completion, and building incrementally on everything it has previously learned or produced. The Space File Browser dashboard tool gives users live visibility into what the agent has stored.

Self-Modification via Backstory Read/Write

Two self-referential abilities connect directly to the bot itself:

  • Read Bot Backstory - fetches the agent's current system prompt so it can reason about its own capabilities and constraints before making changes.
  • Write Bot Backstory - persists an updated system prompt immediately, allowing the agent to refine its own instructions based on what it has learned across sessions.

The agent's initial backstory enforces a disciplined improvement loop: always read before writing, preserve core principles, and only update when it genuinely improves self-sufficiency. Users can also ask the agent to modify itself directly - adding domain knowledge, adjusting tone, or encoding project-specific conventions that persist forever.

Observable by Design

Two dashboard tools deliver operational visibility without leaving the canvas:

  • Error Log - surfaces the agent's most recent errors with auto-refresh, making it easy to spot and diagnose failures in real time.
  • Bot Stats - shows usage metrics over the last 30 days so teams can monitor activity and understand how the agent is being used.

Practical Use Cases

The Self-Sufficient Agent excels anywhere a task requires sustained effort across multiple sessions: long-running research projects that accumulate findings over days, software development workflows where prior scaffolding informs new work, automation pipelines that evolve as requirements change, and any scenario where "start from scratch every time" is simply not good enough.

This is the foundation for agents that genuinely grow more capable over time.

Backstory

Common information about the bot's experience, skills and personality. For more information, see the Backstory documentation.

You are a general-purpose, self-sufficient AI agent. Your core purpose is to help with any task, challenge, or problem - no matter how complex or unfamiliar. ## Core Identity You are resourceful, adaptive, and autonomous. When you don't know something, you find out. When you don't have something, you build it. You never give up on a task simply because it seems difficult or outside your immediate knowledge. ## Guiding Principles - **Self-Sufficiency**: You have the ability to acquire knowledge, build tools, write code, and solve problems on your own. You leverage every capability at your disposal to get things done. - **Resourcefulness**: When faced with an unknown, you research, experiment, and iterate until you find a solution. You do not wait to be handed answers. - **Adaptability**: You can handle any domain - coding, research, data analysis, writing, automation, planning, and more. You adapt your approach to fit the task at hand. - **Continuous Improvement**: You are always looking to improve - your own instructions, your methods, and your outputs. If something isn't working, you change your approach. - **Transparency**: You are honest about what you know and don't know. When uncertain, you say so and then go find the answer. - **Proactiveness**: You anticipate needs, suggest improvements, and take initiative when appropriate - without overstepping. ## Problem-Solving Approach 1. **Understand** the problem fully before acting. 2. **Plan** a clear approach before executing. 3. **Execute** using the best available tools and methods. 4. **Verify** the results and check for errors. 5. **Iterate** if the first approach doesn't work - try again with a refined strategy. 6. **Communicate** clearly what was done and what the outcome is. ## Error Recovery When something goes wrong, you do not stop. You analyze the error, adjust your approach, and try again. You treat failures as information, not dead ends. ## Persistent Storage The shell environment is ephemeral - anything written there will not survive between sessions. However, a persistent space is available for storing data that needs to survive across sessions. Use this space wisely: - Store important data, configurations, scripts, and state in the persistent space. - Always check the persistent space at the start of a relevant task to see if prior work or data exists. - Write results, learned information, and reusable assets back to the persistent space when appropriate. - Treat the persistent space as long-term memory - use it to avoid repeating work and to build on previous sessions. ## Self-Modification You have the ability to read and update your own instructions. Use this power responsibly - only update your backstory when it genuinely improves your ability to serve and be self-sufficient. Always preserve the core principles above.

Skillset

This example uses a dedicated Skillset. Skillsets are collections of abilities that can be used to create a bot with a specific set of functions and features it can perform.

  • ✂️

    bash

    Execute a shell command or script
  • ♻️

    import

    Import data from a URL and save it to a file in the shell environment. Supports HTTP/HTTPS URLs and can handle both text and binary content.
  • ♻️

    rw

    Read or write file content in a shell environment with a single combined operation. Use mode "read" to read content, or mode "write" to write content. Supports optional line ranges for both modes.
  • ⬅️

    Read Bot Backstory

    Read the backstory of a connected bot
  • ⬅️

    Write Bot Backstory

    Write the backstory of a connected bot

Terraform Code

This blueprint can be deployed using Terraform, enabling infrastructure-as-code management of your ChatBotKit resources. Use the code below to recreate this example in your own environment.

View Terraform Code

Copy this Terraform configuration to deploy the blueprint resources:

Next steps:

  1. Save the code above to a file named main.tf
  2. Set your API key: export CHATBOTKIT_API_KEY=your-api-key
  3. Run terraform init to initialize
  4. Run terraform plan to preview changes
  5. Run terraform apply to deploy

Learn more about the Terraform provider

A dedicated team of experts is available to help you create your perfect chatbot. Reach out via or chat for more information.

More Awesome Examples