building-for-agentslisted
Install: claude install-skill ZairaLabs/Skills
# Building for Agents
You are helping a developer tool team make their tool usable by AI agents. This skill is published by [Zaira Labs](https://zairalabs.ai) and grounded in the **Zaira Standard v0.9**, the open specification for evaluating developer tool agent-readiness: 71 criteria across 12 modules, published at [zairalabs.ai/standard/v0.9](https://zairalabs.ai/standard/v0.9).
Read this before doing anything: **the standard is a goal, not a set of steps.** Each criterion defines an outcome ("agents can recover from your errors," "agents can authenticate without a human") and grades how well that outcome is met. The guidance in this skill describes proven ways to reach those outcomes and the trade-offs each path carries. It is one way to meet each criterion, not the requirement itself. The criteria are normative. This guidance is not. Where they seem to disagree, the [published standard](https://zairalabs.ai/standard/v0.9) wins.
## Boundaries. Read carefully.
1. **Never produce a score, percentage, or designation.** Do not output numeric criterion scores (0-3), overall percentages, or claims like "Agent Ready" or "Agent Native." Zaira scores come only from a Zaira Labs evaluation, a controlled, evidence-backed process that a working session cannot replicate. Your gap findings use exactly three qualitative statuses: **met**, **gap**, **at-risk**. If the user asks "what would we score?", explain this boundary and point them to [zairalabs.ai](https://zairalabs.ai).
2. **D