think-cog

**Judgment** A well-scoped, low-friction skill that repackages CellCog’s agent chat mode into an explicit pattern for complex, exploratory thinking. It is best viewed as a structured prompt + workflow convention rather than a deeply engineered autonomous agent. **Strengths** - Excels at ambiguous, high-level problems where outcomes are not known in advance (architecture decisions, strategy, creative direction, decision-making). - Strong documentation and examples make it easy to understand when to use think-cog vs execution-oriented CellCog skills. - Human-in-the-loop design (agent mode, not autonomous teams) reduces automation risks and fits real-world collaborative workflows. **Key Risks / Limitations** - Hard dependency on the separate `cellcog` skill for SDK setup and API calls; if that is misconfigured or changes, this skill will not function. - Provides no additional tools or guarantees beyond the base model + CellCog; quality of reasoning and safety depends heavily on the underlying model and how users steer the conversation. - Not suited for fire-and-forget batch processing, deterministic pipelines, or tasks requiring strict SLAs or reproducibility. **Recommended Scenarios** Use this skill when you want a structured, dialog-based thinking partner for: - Exploring architecture or technical tradeoffs before committing to a design. - Working through business or product strategy where the right path is unclear. - Brainstorming creative directions or campaigns prior to execution. - Debugging complex problems via iterative questioning and hypothesis testing. Avoid it as the primary tool for fully automated, large-scale execution workflows or where you need rigid, auditable decision procedures without human oversight.