Claude Skills by Kastalien-Research

Structured research using Thoughtbox IRCoT (Interleaved Retrieval and Chain-of-Thought). Triggers on: "research", "investigate", "explore", "literature review", "what do we know about".

Analyze completed Thoughtbox sessions to extract patterns, anti-patterns, and learnings for the knowledge graph. This is the learning loop that makes Thoughtbox improve over time. Use after completing a reasoning session, when reviewing past sessions for insights, or when you want to assess reasoning quality. Triggers on "review session", "what did I learn", "extract insights", "analyze reasoning", "session retrospective", or proactively after any significant session completes.

Spawn an Agent Team with a background Thoughtbox analyst teammate. The analyst monitors your reasoning sessions for evolution candidates, contradictions, and generates periodic digests. Use this at the start of any deep Thoughtbox reasoning session, when you're about to record many thoughts and want background analysis. Triggers on "start thoughtbox team", "I want a session analyst", "monitor my reasoning", or at the start of any session where you plan to use Thoughtbox extensively.

Start a Unified Loop Controller session via Ralph Wiggum. Autonomously works through backlog, dispatches agents, and explores via QD when idle.

Thoughtbox-first Ulysses workflow for surprise-gated debugging. Use this when debugging gets uncertain and you need explicit plan, outcome, and reflection discipline without relying on local shell state.

This skill should be used before implementing features, building components, or making changes. It guides exploring user intent, approaches, and design decisions before planning. Triggers on "let's brainstorm", "help me think through", "what should we build", "explore approaches", ambiguous feature requests, or when the user's request has multiple valid interpretations that need clarification.

Evaluate whether an idea is worth implementing by asking structured questions about outcomes, alignment, and opportunity cost. Stage 1 of the development workflow.

Finalize the workflow by reflecting on the process, moving ADRs, closing issues, and preparing for merge. Stage 8 of the development workflow.

Iterate on implementation fixes surfaced by review until all claims are verified. Stage 6 of the development workflow.

Execute an implementation by dispatching multiple sub-agents to parallel Git worktrees. Each builds the entire feature independently, and the best implementation is selected and merged.

Orchestrate the full development lifecycle from ideation through merge. Sequences 8 stages with gates, dispatches to stage skills, and maintains workflow state.

Capture learnings from the workflow into reusable documentation. Stage 7 of the development workflow.

Transform spec (with frontmatter claims) into an implementation plan with task decomposition and sub-agent assignment. Stage 3 of the development workflow.

Verify implementation claims by dispatching review agents. Stage 5 of the development workflow.

Execute the implementation plan by dispatching sub-agents for each task. Stage 4 of the development workflow.

Comprehensive prompt engineering reference for Claude Opus 4.6. Use this skill whenever you are writing, reviewing, or optimizing prompts for Claude Opus 4.6 (or Sonnet 4.6 / Haiku 4.5). Covers foundational clarity principles, output formatting control, tool use, adaptive thinking, agentic system design, multi-context-window workflows, and migration guidance from earlier models. Trigger this skill for any task involving system prompt authorship, agent harness design, multi-step pipeline const...

Hypothesis-Driven Development — treat ADRs as the source of truth, not code. Form testable hypotheses, stage ADRs, implement to test them, validate against predictions, then accept or reject based on evidence. Prevents repeated mistakes by documenting failures. Enables agent-to-agent collaboration through explicit architectural reasoning. This is the canonical HDD entry point — it replaces the thin router at .claude/commands/hdd/hdd.md.