Human Mode

Non-technical stakeholders often struggle to provide clear answers when developers or AI agents present complex technical choices. This lack of clarity leads to project stalls or misinformed decisions that are expensive to reverse. This skill bridges the gap by translating technical requirements into plain-English trade-offs while maintaining full technical rigor in the actual implementation.

What it does

• Translation layer rewrites technical queries into plain-English explanations without hiding the underlying technical terms.
• Level profiling adapts the complexity of language based on the user's coding experience, from "never coded" to "some coding."
• Decision logging maintains a DECISIONS.md file tracking what was chosen, why it was chosen, and the cost of changing it later.
• Risk flagging highlights specific choices that impact security, incur monthly costs, or risk data loss.
• Recommendation engine identifies the most logical technical path and explains the reasoning in everyday terms.

How it works

1. Profiling: The agent asks a one-time setup question to determine the user's technical literacy and storage preferences.
2. Translation: Before any technical choice, the agent presents the real technical name followed by a plain-language analogy and a practical consequence.
3. Mapping: The user responds in natural language (e.g., "the free one"), and the agent maps this back to the specific library or framework.
4. Logging: Every choice is documented in the project root to prevent repetitive questions or architecture contradictions.

Frameworks & tools

The skill is designed for Claude Code and the Claude app. It manages local profiles in ~/.claude/ and maintains documentation in Markdown format for any development project.

Why this beats prompting it yourself

Manual prompting often results in the AI either being too technical for the user to understand or too vague for the developer to implement. This skill enforces a consistent structure that keeps technical names visible while providing the exact business consequences of every choice.

Use cases

• Infrastructure choices such as explaining the difference between SQLite and PostgreSQL to a first-time founder.
• Architecture approvals like explaining why a specific auth provider is recommended over building a custom one.
• Cost management for projects where the client needs to approve any service that triggers a recurring bill.

Known limitations

Does not simplify the actual code output or technical documentation. Relies on filesystem access to persist profiles across different sessions in Claude Code.