如果你觉得AI的分析能力，判断能力不如人类，那可能是你没教会它什么场景该用什么样的刀。Mindthus帮助AI Agent发现问题，定义问题并解决问题，把混沌问题拆解明白，带着目标去加深思考深度，输出厚度。
Skill Claude CodeCodex
Use when a problem is unclear, noisy, repeatedly reworked, or too large to execute directly; also use when a task list looks structured but may not be executable, verifiable, or tied to a falsifiable problem.
Skill Claude CodeCodex
Use when an ambiguous qualitative judgment needs Extreme Deduction and optional Scenario Projection: push variables to extremes, build a structural coordinate system, diagnose malformed propositions, then project concrete scenarios when needed.
Skill Claude CodeCodex
Use when a plausible mainline or strategic direction meets a concrete carrier/proxy, concentrated scarce-resource exposure, or continue/commit/hold/exit/switch decision under path volatility, costs, delays, fragility, or counter-forces; not for method-reference evidence review.
Skill Claude CodeCodex
Use for long-term system-efficiency versus local-advantage direction pressure, especially old paradigm vs scalable system or external system efficiency vs local/internal advantage; when concrete carrier/exposure/path action is also present, use as a support lens with MPG rather than sole owner.
Skill Claude CodeCodex
Use when an AI agent needs an OKR-Runtime: a Mission needs script-driven task state, acceptance evidence, decision hooks, human-in-loop authority, or Mission-relative addition, subtraction, selection, closure, and recovery.
Skill Claude CodeCodex
Use as a value-directed strengthening loop for bounded AI artifacts that need clearer judgment, evidence, trade-offs, handoff, reuse, or action value.
Skill Claude CodeCodex
Use only for fact-sufficient hard-judgment routing among Mindthus lenses; structural, strategic, path, control, artifact, problem-definition, Mission-state, or allocation ambiguity. Never load for an ordinary request lacking facts or resource bounds to compare candidates; ask directly. SRA only after multiple…
Skill Claude CodeCodex
Use when a workflow, script, schema, agent loop, evidence gate, or review step may be controlling the wrong part of the work; especially when clean structure may be freezing uncertain truth or hiding thin judgment.
Skill Claude CodeCodex
Use directly when multiple already-judgeable actions or bundles compete for one named scarce resource; choose the current floor and next meaningful tranche. Default Lite; use Full for explicit bundles, multiple resource constraints, fixed thresholds, or major commitments. Do not use with missing comparison…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: