Disclaimer — mimicode

AI Output Accuracy

mimicode uses large language models (LLMs) to generate responses. LLMs can produce output that is inaccurate, incomplete, outdated, or confidently wrong — a phenomenon commonly called hallucination. This includes but is not limited to:

Incorrect code logic or syntax
Wrong library versions, deprecated APIs, or made-up function names
Flawed analysis or debugging conclusions
Explanations that sound authoritative but are factually incorrect

Do not treat AI-generated output as authoritative. Always verify, test, and review independently before acting on it.

Code Quality and Security

Code produced by the agent has not been reviewed, tested, or audited by us. It may:

Contain bugs, logic errors, or performance issues
Introduce security vulnerabilities such as injection flaws, insecure defaults, or improper input handling
Fail to handle edge cases, errors, or concurrent access correctly
Be unsuitable for your specific environment, language version, or dependencies

Always review and test AI-generated code thoroughly before deploying it to production or using it with sensitive data. mimicode is a tool to assist your judgment, not replace it.

File Modifications Are Immediate

When mimicode uses the write or edit tool, those changes are applied immediately to your filesystem. The shadow-git undo feature (:undo in REPL mode) can revert recent turns, but it is not a complete backup solution and may not be recoverable in all failure scenarios.

Important: Maintain independent backups of important files and repositories. Do not rely solely on mimicode's undo feature to protect critical work.

Shell Command Execution

The bash tool executes real shell commands on your machine with your user's permissions. Commands run by the agent can:

Delete or overwrite files and directories
Modify system or application configuration
Install, update, or remove software
Make outbound network requests
Incur cloud or infrastructure costs
Start or stop running services

Use --confirm mode (or set MIMICODE_CONFIRM=1) to require explicit approval before every bash, write, and edit call. Read-only tools never require confirmation.

For production or sensitive systems: always run with --confirm enabled and review every proposed action before approving it.

API Costs

Every task you run incurs API usage billed to your Anthropic account. Complex, multi-step tasks with many tool calls can consume substantial token budgets — particularly when using high-thinking-budget model settings.

We are not responsible for any API charges incurred through your use of mimicode. Monitor your usage in the Anthropic console. You can limit spending by setting lower thinking budgets via the MIMICODE_THINKING environment variable or by capping MIMICODE_MAX_STEPS.

No Professional Advice

Nothing generated by mimicode constitutes legal, security, financial, medical, or other professional advice. Do not rely on AI output as a substitute for qualified professional expertise in regulated, safety-critical, or high-stakes domains.

AI-generated security analysis should never replace a professional penetration test or security audit. AI-generated legal analysis is not legal advice. AI-generated medical information is not medical advice.

Limitation of Liability

TO THE FULLEST EXTENT PERMITTED BY APPLICABLE LAW, MIMICODE AND ITS AUTHORS EXPRESSLY DISCLAIM ALL LIABILITY FOR ANY DAMAGES ARISING FROM:

Inaccurate, incomplete, or harmful AI-generated output
File or system modifications made by the agent
Shell commands executed by the agent
Data loss, corruption, or unintended deletion
Security vulnerabilities introduced by AI-generated code
API costs or billing disputes with Anthropic
Any other direct, indirect, incidental, or consequential loss

YOUR USE OF MIMICODE IS ENTIRELY AT YOUR OWN RISK.

Contact

Questions or concerns? Open an issue or discussion at github.com/trymimicode/mimicode-go/issues.