<!-- Made by Alejandro Apodaca Cordova (apoapps.com) -->

OpenCode Result Handling

This skill governs how Claude processes responses from OpenCode to maximize token savings while maintaining quality.

Core principle

Claude delegates exploratory, analytical, and review work to OpenCode. Claude then validates and synthesizes the response instead of regenerating the analysis from scratch. This saves Claude tokens by letting the configured model do the heavy lifting.

When presenting OpenCode output

• Preserve the structure: findings, recommendations, and code suggestions.
• Keep file paths and line numbers exactly as reported.
• If OpenCode provides code snippets, present them as-is unless they contain obvious errors.
• Preserve severity ratings and confidence levels.
• If a fallback model was used, note it briefly: "Note: used fallback model X (primary Y was unavailable)."

Validation rules

Claude MUST validate OpenCode responses before presenting to the user:

1. Sanity check: Does the response address the actual question?
2. Code correctness: Are code suggestions syntactically valid?
3. File references: Do referenced files/functions actually exist? (Spot-check 1-2 if suspicious.)
4. Completeness: Did it miss anything obvious?
5. Hallucination detection: Flag if OpenCode references APIs, functions, or patterns that don't exist in the codebase.

Validation cost budget

• Spend at most ~200 tokens on validation commentary.
• If OpenCode's answer is solid, just present it with a one-line "Validated by Claude" note.
• If issues found, add a brief "Claude notes" section at the end.

Token-saving patterns

• Don't re-explain what OpenCode already explained well. Just say "OpenCode's analysis is accurate" and add only delta.
• Don't rewrite code that OpenCode wrote correctly. Present it as-is.
• Don't add hedging ("I think", "it seems like"). Be direct.
• Batch validations: If multiple findings, validate in one pass, not per-finding.

When OpenCode fails

• If OpenCode returned an error or empty response, say so briefly.
• Do NOT generate a substitute answer to fill the gap.
• Suggest the user try again or use Claude directly for this specific question.
• If setup/auth is required, direct to /opencode:setup.
• If the primary model failed but a fallback succeeded, note which model was used.

This plugin is primary — Codex plugin is fallback

OpenCode CLI gives you access to ALL providers in one place, including Codex models (openai/gpt-5-codex, openai/gpt-5.1-codex, github-copilot/gpt-5.1-codex, etc.). This means:

• Use this plugin first. It covers lightweight models (MiniMax) AND heavy models (Codex) through a single interface.
• Codex plugin is a fallback for when you need Codex-specific features (app-server protocol, resume threads, write-capable runs).
• Why? OpenCode unifies model access — no need for separate Codex CLI auth, separate state, or separate workflows. One plugin, all models.

When to fall back to Codex plugin:

• Write-capable runs (OpenCode here runs read-only for safety).
• Resumable threads (Codex app-server supports thread resume).
• When OpenCode CLI itself is down or misconfigured.

CRITICAL

After presenting review findings from OpenCode, STOP. Do not auto-fix. Ask the user which issues they want fixed before touching any file.