Reasoning Workflow
How to use the reasoning agent and skills for structured thinking, problem solving, creative ideation, and learning.
Overview
The reasoning system provides structured cognitive capabilities through 4 composable skills coordinated by 1 agent.
+---------------------------------------------------------------+
| REASONING |
+---------------------------------------------------------------+
| AGENT |
| +---------------------+ |
| | rsn-problem-solver | Coordinates perceive --> think loop |
| +---------------------+ |
+---------------------------------------------------------------+
| SKILLS |
| +------------------+ +------------------+ |
| | rsn-perceiving- | | rsn-reasoning- | |
| | information | | problems | |
| | 6 modes | | 6 modes | |
| +------------------+ +------------------+ |
| +------------------+ +------------------+ |
| | rsn-learning- | | rsn-creating- | |
| | outcomes | | ideas | |
| | 5 modes | | 10+ techniques | |
| +------------------+ +------------------+ |
+---------------------------------------------------------------+
Cognitive Cycle
ENVIRONMENT
|
v
rsn-perceiving-information (What's happening?)
|
v
rsn-reasoning-problems (What does it mean?)
|
+-----------+-----------+
| |
v v
rsn-creating-ideas ACTION
(What could we do?) (Execute)
|
v
rsn-learning-outcomes
(What did we learn?)
Skill Selection
Decision Tree
What is the primary task?
+-- GATHER INFORMATION
| +-- Don't know what to look for? --> perceiving.scanning
| +-- Need deep understanding? --> perceiving.focusing
| +-- Too much information? --> perceiving.filtering
| +-- Need to verify a claim? --> perceiving.triangulating
| +-- Need ongoing awareness? --> perceiving.monitoring
| +-- Multiple signals to combine? --> perceiving.synthesizing
|
+-- REASON THROUGH SOMETHING
| +-- Have goal to achieve? --> reasoning.causal
| +-- Need to explain observation? --> reasoning.abductive
| +-- Have instances, need rule? --> reasoning.inductive
| +-- Novel situation, recall similar? --> reasoning.analogical
| +-- Conflicting options/tensions? --> reasoning.dialectical
| +-- Evaluating "what if"? --> reasoning.counterfactual
|
+-- GENERATE IDEAS
| +-- Need breakthrough? --> creating-ideas (First Principles)
| +-- Rapid ideation? --> creating-ideas (SCAMPER)
| +-- Stuck in pattern? --> creating-ideas (Lateral Thinking)
| +-- Evaluate idea quality? --> creating-ideas (AUDIT mode)
|
+-- LEARN FROM EXPERIENCE
| +-- Action failed, need quick fix? --> learning.single-loop
| +-- Pattern of failures? --> learning.double-loop
| +-- Experience complete? --> learning.reflection
| +-- Belief needs testing? --> learning.experimentation
| +-- Predictions miscalibrated? --> learning.calibration
|
+-- SOLVE A PROBLEM (multi-step)
+--> rsn-problem-solver agent (coordinates perceive --> think loop)
Trigger Phrases
| If you hear... | Use this |
|---|---|
| "What's out there?" | perceiving.scanning |
| "Tell me more about X" | perceiving.focusing |
| "Too much noise" | perceiving.filtering |
| "Is this true?" | perceiving.triangulating |
| "How do we do X?" | reasoning.causal |
| "Why did X happen?" | reasoning.abductive |
| "I keep seeing X" | reasoning.inductive |
| "This is like..." | reasoning.analogical |
| "On one hand... on the other..." | reasoning.dialectical |
| "What if we had done X?" | reasoning.counterfactual |
| "Brainstorm options" | creating-ideas |
| "What did we learn?" | learning.reflection |
| "Why does this keep happening?" | learning.double-loop |
| "Figure out why..." | rsn-problem-solver |
Composition Patterns
Skills compose into pipelines. The output of one skill becomes input to the next.
Perceive then Think
Most common pattern for analysis tasks.
perceiving.X --> reasoning.Y --> output
Example: scanning --> focusing --> triangulating --> synthesizing --> dialectical --> decision
Think then Act then Learn
For tasks where inputs are already known.
reasoning.X --> action --> learning.Y
Example: abductive --> fix --> single-loop
Create then Think
For tasks needing novel solutions.
creating-ideas --> reasoning.dialectical --> reasoning.causal
Example: ideate --> evaluate trade-offs --> plan execution
Full Cycle
For complete operational workflows.
perceiving --> reasoning --> acting --> learning --> (back to perceiving)
Workflow Patterns
Research to Decision
perceiving.scanning --> perceiving.focusing --> perceiving.triangulating
--> perceiving.synthesizing --> reasoning.dialectical --> decision
Use when: Strategic question requires investigation before deciding.
Incident Response
perceiving.monitoring (trigger) --> reasoning.abductive (diagnose)
--> learning.single-loop (fix) --> [if pattern] learning.double-loop
Use when: Something is broken and needs fixing.
Hypothesis Validation
reasoning.abductive (hypothesis) --> validation criteria
--> learning.experimentation --> learning.calibration --> decision
Use when: Belief needs validation before commitment.
Retrospective
gather data --> learning.reflection --> learning.single-loop
--> [if pattern] learning.double-loop --> recommendations
Use when: Learning from completed project or incident.
Thread Integration
Reasoning skills map to the causal flow thread structure:
| Thread Stage | Reasoning Skill |
|---|---|
| 1-input.md | rsn-perceiving-information (gather context) |
| 2-hypothesis.md | rsn-reasoning-problems (form hypothesis) |
| 3-implication.md | rsn-reasoning-problems.causal (derive implications) |
| 4-decision.md | rsn-reasoning-problems.dialectical (evaluate options) |
| 5-actions.md | Execution (not reasoning) |
| 6-learning.md | rsn-learning-outcomes.reflection (capture learnings) |
Reference Card
rsn-perceiving-information Modes
| Mode | Use When | Key Output |
|---|---|---|
| scanning | Need landscape awareness | Signals + focus candidates |
| focusing | Need deep understanding | Findings + implications |
| filtering | Information overload | Prioritized items |
| triangulating | Need to verify claims | Confidence-rated claims |
| monitoring | Need ongoing awareness | Alerts when threshold crossed |
| synthesizing | Need to integrate signals | Coherent picture |
rsn-reasoning-problems Modes
| Mode | Use When | Key Output |
|---|---|---|
| causal | Executing, planning, procedures | Path from current to goal |
| abductive | Diagnosing, risk identification | Ranked hypotheses |
| inductive | Extracting patterns, summarizing | Pattern with confidence |
| analogical | Applying knowledge from similar domain | Transferred insight |
| dialectical | Resolving conflicts, deciding, evaluating | Synthesis or choice |
| counterfactual | Evaluating alternatives | Comparison of paths |
rsn-learning-outcomes Modes
| Mode | Use When | Key Output |
|---|---|---|
| single-loop | Fixing immediate problem | Correction + verification |
| double-loop | Questioning the frame | Reframed assumptions |
| reflection | Extracting from experience | Insights + artifacts |
| experimentation | Testing hypothesis | Validated/invalidated + confidence |
| calibration | Adjusting predictions | Calibration rules |
rsn-creating-ideas Techniques
| Technique | Use When | Key Output |
|---|---|---|
| First Principles | Need breakthrough | Rebuilt from fundamentals |
| SCAMPER | Rapid ideation | Systematic variations |
| Lateral Thinking | Stuck in pattern | Unexpected connections |
| Analogical Transfer | Need cross-domain insight | Borrowed solutions |