If your question smells like something Stack Overflow answered in 2013, don’t summon a team meeting like it’s an emergency. Ask an AI Assistant first: it won’t judge you (and it doesn’t have meetings to attend)
This table defines the situations in which an AI assistant should (or should not) be your first stop. The aim is to save team time while ensuring Candido’s sensitive information never leaves our perimeter.
| Scenario | AI Assistant? | Rationale |
|---|---|---|
| Pure language‑syntax or computer‑science questions (“How do I sort a list of tuples by the second element in Python?”) | ✅✅✅ Yes, mandatory! | No project context needed; answer exists in public documentation. |
| Library / tooling configuration (“Configure pytest‑xdist for parallel tests”) | ✅✅✅ Yes, mandatory! | Well‑documented topic, low IP exposure. |
| Common UI phrases for web platforms (How to phrase a tooltip for a disabled button?) | ✅✅✅ Yes, mandatory! | Frequently reused content; no sensitive data or project-specific context required. |
| Boilerplate generation (Dockerfile, GitHub Action YAML, basic API stub) | ✅ Yes, after sanitising | Routine patterns; minimal proprietary logic. |
| Small, self‑contained bug fix where ≤30 lines of code suffice to show the issue | ✅ Yes, but double‑check output | Limited snippet keeps proprietary logic private. |
| Architecture discussion, proprietary algorithms, data schema, or strategic roadmap | 🚫🚫🚫 No | High risk of exposing Candido IP. |
| Large code blocks (>30 lines), config files, or real datasets | 🚫 No | Easy to leak sensitive details unintentionally. |
| Debug sessions that require full stack traces with internal paths or identifiers | 🚫 No | Stack traces often reveal too much context. |
<aside> 💡
Rule of thumb: If the answer is something an AI assistant (like ChatGPT) can confidently provide — such as syntax help, configuration tips, or solving generic bugs — always start there. It’s strongly discouraged to interrupt a teammate for something you could have resolved in seconds with a well-defined prompt.
Only reach out to a colleague when the question truly requires Candido-specific context, like internal architecture, proprietary logic, or decisions that involve team alignment.
</aside>
AI Assistance for Candido developers - an Handbook