Menu
Ask Neo for semantic reasoning and code suggestions over the current codebase. Use for general questions, code suggestions, or architectural guidance backed by Neo's persistent memory.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Get Neo's architectural guidance for design decisions. Trade-off analysis for choices like microservices vs monolith, sync vs async, event-driven vs request-response — with persistent memory of how similar decisions played out.
Ask Neo to help debug intermittent, complex, or hard-to-reproduce issues. Particularly useful for race conditions, memory issues, distributed-systems bugs, and cases where the symptom is not the root cause.
Ask Neo for optimization suggestions on a function, file, or hot path. Targets algorithmic improvements, redundant work, allocation/hot-loop issues — not micro-style.
Ask Neo to extract a reusable pattern from a piece of code, or to find existing patterns in the codebase that match a description. Useful for codifying conventions and finding duplicated logic.
Get Neo's code review with semantic pattern matching. Focuses on security vulnerabilities, edge cases, error handling, and performance issues across the target file or module.
Sync current branch with base branch. Codex (you) resolves conflicts intelligently, not automated scripts.
| name | neo |
| description | Ask Neo for semantic reasoning and code suggestions over the current codebase. Use for general questions, code suggestions, or architectural guidance backed by Neo's persistent memory. |
When the user invokes this skill ($neo <question or task>), do the following:
Verify Neo is installed. Run neo --version once. If the command is missing, tell the user: "Neo CLI not installed. Run pip install neo-reasoner[openai] and set OPENAI_API_KEY, then retry." Stop.
Gather context. Use your file-reading tools to collect the most relevant files for the user's question (typically 1–5 files). Include the user's full question text.
Invoke Neo via stdin heredoc. Allow up to 5 minutes — Neo runs multi-agent reasoning across LLM calls.
neo --mode advise <<'QUERY'
<restate the user's question here, plus any short context excerpts>
QUERY
Parse Neo's output. Neo returns four structured sections: CONFIDENCE, PLAN, SIMULATIONS, CODE SUGGESTIONS. Each suggestion carries its own confidence score.
Present results to the user. Lead with Neo's overall confidence and the highest-confidence suggestion. Quote Neo's reasoning verbatim where it's load-bearing. If confidence is below 0.7, flag it explicitly.
advise mode: it may retrieve established memory but does not create candidates or update durable memory. Use neo --mode learn only when the user intends to contribute outcome evidence.$neo-review, $neo-optimize, $neo-architect, $neo-debug, $neo-pattern).