ichendev/prompt-optimizer

Audit and subtractively clean Skills, AGENTS.md, and coding-agent scaffolding for GPT-6 Astra. Use when Astra stops early, loads the wrong skills, over-reads, over-tests, over-asks, or when migrating Sol-era harness instructions—not for rewriting product prompts (use optimize-gpt-6-astra-prompts).

Clarify and rewrite prompts for Claude Fable 5. Use when optimizing, migrating, debugging, or designing a Fable 5 / Fabel prompt (including Fable/Fabel 提示词优化)—not for Claude Opus 5.

Clarify and rewrite prompts or agent harnesses for Claude Opus 5 (claude-opus-5). Use when optimizing, migrating from Opus 4.8, or tuning Opus 5 behaviors such as long responses, over-narration, scope expansion, eager subagents, or over-verification (including Opus 5 提示词优化)—not for Claude Fable 5.

Clarify and rewrite prompts or agent harnesses for Claude Sonnet 5.5 (claude-sonnet-5-5). Use when optimizing, migrating from Sonnet 5, or tuning Sonnet 5.5 behaviors such as effort, initiative/scope, between_tools, JSON accuracy, progress updates, or low-effort verification (including Sonnet 5.5 提示词优化)—not for Opus/Fable prompt rewrites or bare model picking (use select-claude-models).

Clarify and rewrite prompts for DeepSeek-V4 Pro and Flash. Use when optimizing, debugging, migrating, or designing a DeepSeek-V4 prompt (including DeepSeek / DeepSeek-V4 提示词优化); keep reasoning modes and special tokens outside prompt prose.

Clarify and rewrite prompts for the Gemini model family. Use when optimizing, debugging, migrating, or designing a Gemini prompt (including Gemini 提示词优化), especially multimodal or long-context cases.

Clarify and rewrite prompts for the GLM / 智谱 model family. Use when optimizing, debugging, migrating, or designing a GLM prompt (including GLM 提示词优化 or 智谱提示词优化).

Clarify and rewrite prompts for GPT-5.6 Sol and the GPT-5.6 family. Use when optimizing, debugging, migrating, or designing a GPT-5.6 prompt (including GPT-5.6 提示词优化)—not for GPT-6 Astra (use optimize-gpt-6-astra-prompts or audit-gpt-6-astra-skills).

Clarify, migrate, and rewrite product prompts for GPT-6 Astra using OpenAI's latest-model guidance. Use when optimizing, debugging, migrating, or designing a GPT-6 Astra / gpt-6-astra prompt—not for Skills/AGENTS.md harness cleanup (use audit-gpt-6-astra-skills).

Clarify and rewrite prompts for the Kimi model family. Use when optimizing, debugging, migrating, or designing a Kimi prompt (including Kimi 提示词优化).

Recommend Claude model choices and multi-agent pairings (Opus 想 / Sonnet 干活, advisor-executor, orchestrator-worker). Use when picking among Opus 5.5, Sonnet 5.5, Fable, or Haiku, or designing Claude cost-versus-intelligence routing—not for rewriting a single model's prompt text.