Files
gh-synapseradio-thinkies-pl…/commands/reason.md
2025-11-30 08:59:09 +08:00

3.5 KiB

name, description
name description
reason Comprehensive reasoning session with full introspection lifecycle and persistent state management

/reason

A structured reasoning session with mandatory note-keeping, working memory management, and reflection throughout the entire process.

Execution Protocol

When this command is invoked, you follow this protocol precisely.

Initialize

Session Naming: Propose a natural, descriptive session name based on the problem (e.g., "auth-architecture-decision", "performance-root-cause"). Announce this name to the user - it unifies all artifacts created during the session.

Check Prior Sessions: Ask the user whether they want to reference prior sessions for context. If they confirm, invoke Skill("introspect-reference-notes") to list recent sessions and search for related work. If a relevant prior session exists, offer to continue from it. When continuing, use the --continue flag when saving notes to create a new session that builds on the prior content. If the user declines, proceed directly to context setup.

Context Setup: Invoke Skill("introspect-working-memory") to externalize the initial problem context. This establishes what you're holding in active focus as the session begins.

Frame Phase

Apply Skill("analysis-decompose") and Skill("metacognitive-knowledge-check") to understand the problem space.

After framing, invoke Skill("introspect-keep-notes") to record key insights from decomposition, what is known vs unknown, and initial constraints and goals.

Explore Phase

Apply generative skills as appropriate: Skill("exploration-wonder"), Skill("exploration-branch"), Skill("analysis-alternatives").

After exploring, invoke Skill("introspect-keep-notes") to record options generated, branches identified, and creative possibilities worth considering.

Analyze Phase

Apply analytical skills: Skill("analysis-evidence-check"), Skill("techniques-systems-thinking"), Skill("analysis-logic-trace"), Skill("decision-premortem") as the problem warrants.

After analysis, invoke Skill("introspect-keep-notes") to record evaluation of each option, risks and failure modes identified, and dependencies or interactions discovered.

Decide Phase

Apply decision skills: Skill("decision-weigh-options"), Skill("decision-priority-matrix").

After deciding, invoke Skill("introspect-keep-notes") to record the decision and its rationale, alternatives considered and why rejected, and conditions that would warrant revisiting.

Reflect Phase

Invoke Skill("introspect-session-reflect") to evaluate the reasoning process itself. Persist the reflection with the session name.

Invoke Skill("introspect-self-observation") to log which skills were used during the session, building the usage analytics.

Invoke Skill("introspect-working-memory") to save final context state.

Complete

Report to the user:

  • The session name for future reference
  • A summary of the decision or conclusion reached
  • Location of persistent artifacts: .claude/.thinkies/.notes/ and .claude/.thinkies/.reflections/

These directories are automatically gitignored. The user can reference this session later or continue from it in a future /reason invocation.

When to Use This Command

Use /reason for significant decisions or complex problems where you want full traceability. The command creates a complete record of the reasoning process that persists beyond the session.

For lighter reasoning needs where full instrumentation isn't necessary, use /reason-session instead.