Session Review (Learning Capture)
Review the current session and capture learnings for future sessions.
Input: $ARGUMENTS (optional: specific area to focus review on)
Purpose
Sessions contain valuable learning opportunities that are lost if not captured. This procedure systematically extracts learnings to improve future sessions.
Principle: Every session should teach. Capture what worked, what did not, and what surprised you.
Step 1: Session Summary
Summarize what happened in this session:
| Aspect | Description |
|---|---|
| User’s initial input | [what they brought] |
| Type | GOAL / PROBLEM / QUESTION / DECISION / SITUATION / FEELING |
| Procedures invoked | [list of /procedures used] |
| Outcome | [what was achieved] |
| Session duration | [if known] |
Step 2: What Worked
Identify approaches that were effective:
| What Worked | Why It Worked | Reusable? |
|---|---|---|
| [approach 1] | [reason for success] | YES / NO |
| [approach 2] | [reason for success] | YES / NO |
| [approach 3] | [reason for success] | YES / NO |
Questions to ask:
- Which procedure generated the most value?
- What questions unlocked insight?
- What framing helped the user?
- What would I do again?
Step 3: What Did Not Work
Identify approaches that were ineffective:
| What Did Not Work | Why It Failed | Avoidable? |
|---|---|---|
| [approach 1] | [reason for failure] | YES / NO |
| [approach 2] | [reason for failure] | YES / NO |
Questions to ask:
- Where did I go off track?
- What procedure was wrong for this situation?
- What assumption was incorrect?
- What would I skip next time?
Step 4: Surprises
Identify unexpected findings:
| Surprise | Implication |
|---|---|
| [unexpected finding 1] | [what it means for future] |
| [unexpected finding 2] | [what it means for future] |
Questions to ask:
- What was harder than expected?
- What was easier than expected?
- What did the user teach me?
- What did I not anticipate?
Step 5: Procedure Gaps
Identify missing or inadequate procedures:
| Gap Type | Description | Proposed Solution |
|---|---|---|
| Missing procedure | [needed but does not exist] | [what it would do] |
| Inadequate procedure | [exists but insufficient] | [what needs improvement] |
| Wrong chaining | [procedures invoked in wrong order] | [correct order] |
| Overly complex | [procedure too heavy for situation] | [simpler alternative] |
Step 6: Convergent Validation of Learnings
For each potential learning, apply convergent validation:
| Learning | Grounded? | Fixed Point? | Convergent? | Practical? | Score |
|---|---|---|---|---|---|
| [learning 1] | [O/T/D marker?] | [stable under re-analysis?] | [multiple paths lead here?] | [actionable?] | [0-4] |
| [learning 2] | [O/T/D marker?] | [stable under re-analysis?] | [multiple paths lead here?] | [actionable?] | [0-4] |
Decision:
- 4 checks pass: Accept with high confidence
- 3 checks pass: Accept with medium confidence
- 2 checks pass: Flag for review (preliminary)
- <2 checks pass: Reject (do not capture)
Step 7: Capture Validated Learnings
For each accepted learning, format for storage:
LEARNING CAPTURE:
ID: [auto-generated]
TYPE: what_worked | what_didnt | surprise | insight | pattern | procedure_gap
CONTENT: [the learning in one sentence]
CONTEXT: [when/where this applies]
DOMAIN: goal | problem | question | decision | situation | feeling | general
CONFIDENCE: high | medium | preliminary
EVIDENCE: [supporting evidence from session]
GROUNDING: [O: observed] | [T: tested] | [D: derived]Execute: Store using python scripts/session_learning.py capture [type] [content] [context]
Step 8: Review Previous Learnings
Check if any previous learnings were validated or contradicted in this session:
| Previous Learning | Session Evidence | Update |
|---|---|---|
| [learning ID] | [what happened] | VALIDATE / CONTRADICT / NO CHANGE |
Execute: Update using python scripts/session_learning.py validate [learning_id] [true/false]
Step 9: Test Guess Outcomes
For any recommendations or actions taken based on guesses in this session:
| Recommendation | Based On Guess | Outcome | Update Confidence |
|---|---|---|---|
| [what was recommended] | [underlying guess] | WORKED / DIDNT_WORK / UNKNOWN | [increase/decrease/no change] |
Questions to ask:
- Did the recommended action achieve its goal?
- Did any assumption prove false?
- Did any guess prove correct?
Capture outcomes:
- If WORKED:
capture_learning(type="what_worked", content="[recommendation] based on [guess]") - If DIDNT_WORK:
capture_learning(type="what_didnt", content="[recommendation] failed because [reason]")
Step 10: Check for Contradictions
Review any potential conflicts detected:
Execute: python scripts/session_learning.py conflicts
For each conflict:
| Learning A | Learning B | Resolution |
|---|---|---|
| [ID] [content] | [ID] [content] | KEEP_A / KEEP_B / KEEP_BOTH / NEEDS_MORE_DATA |
Resolve: Archive the contradicted learning using validate with false
Output Format
## SESSION REVIEW
### Summary
- Input: [user's input]
- Type: [GOAL/PROBLEM/etc]
- Procedures: [list]
- Outcome: [result]
### What Worked
1. [learning] - [why]
2. [learning] - [why]
### What Did Not Work
1. [learning] - [why]
### Surprises
1. [finding] - [implication]
### Procedure Gaps
1. [gap] - [proposed solution]
### Learnings Captured
[list of learning IDs captured]
### Previous Learnings Updated
[list of learning IDs validated/contradicted]
### Recommendations for Future
1. [recommendation]
2. [recommendation]Integration Points
At session end: Invoke /ssr to capture learnings.
At session start: Retrieve relevant learnings with:
python scripts/session_learning.py context [domain]Periodically: Review learning stats with:
python scripts/session_learning.py statsExecution Checklist
- Session summarized
- What worked identified (aim for 2-3)
- What did not work identified (aim for 1-2)
- Surprises noted
- Procedure gaps identified
- Learnings validated through convergent checks
- Accepted learnings captured to storage
- Previous learnings updated if applicable
Execute now: Review this session and capture learnings.