Agent skill

reflect

Reflect on a session to identify learnings, friction points, and improvements. Captures valuable insights for future sessions and system evolution.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/reflect-kynetic-ai-kynetic-spec

SKILL.md

Session Reflection

Structured reflection using the @session-reflect workflow.

Quick Start

bash
# Start the reflection workflow
kspec workflow start @session-reflect

# Advance through steps (workflow will guide you)
kspec workflow next --notes "your notes..."

Workflow Overview

The reflection workflow has 6 steps:

  1. What Worked Well - Identify effective practices
  2. Friction Points - Where things were harder than needed
  3. Check Coverage - Search specs/tasks/inbox for existing tracking
  4. Propose Improvements - Concrete ideas for untracked friction
  5. Discussion - Present to user, get approval one at a time
  6. Capture - Add approved items to inbox/observations

Use kspec workflow show to see current progress.

Step Details

Step 1: What Worked Well

Identify practices that were effective:

  • Workflows that flowed smoothly
  • Tools/commands that helped
  • Communication patterns that kept alignment
  • Decisions that proved correct

Be specific - "categorizing items first" not "good planning"

Step 2: Friction Points

Identify where things were harder than needed:

  • Repetitive manual steps
  • Missing commands or options
  • Context loss or re-explanation
  • Workarounds used

Focus on systemic issues, not one-off mistakes

Step 3: Check Existing Coverage

Before proposing improvements, search ALL sources:

bash
kspec search "<keyword>"  # Searches specs, tasks, AND inbox

For each friction point, note if it's:

  • Already tracked - reference the existing item/task
  • Partially covered - note what's missing
  • Not tracked - candidate for capture

Exit criteria: Must have searched specs, tasks, and inbox.

Step 4: Propose Improvements

For untracked friction, propose concrete improvements:

  • What it would do
  • How it would help
  • Rough scope (small/medium/large)

Step 5: Discussion

Present findings to user. Ask one at a time about each improvement:

  • Is this worth capturing?
  • Any refinements to the idea?
  • Related ideas from user perspective?

Step 6: Capture

Use appropriate destination:

bash
# Actionable improvements (future work)
kspec inbox add "Description" --tag reflection --tag <area>

# Friction patterns (systemic issues)
kspec meta observe friction "Description"

# Success patterns (worth replicating)
kspec meta observe success "Description"

# Open questions
kspec meta question add "Question?"

Where to Capture What

What you found Where to put it Why
Actionable improvement idea inbox add Will become a task eventually
Friction pattern (systemic) meta observe friction Informs process improvement
Success pattern meta observe success Worth documenting/replicating
Open question needing research meta question add Track during session
Bug or specific fix needed task add Ready to implement

Reflection Prompts

Use these during steps 1-2:

Process: What pattern did I repeat 3+ times? What workarounds did I use? Tools: What command/flag did I wish existed? Communication: Where was the user surprised? What should I have asked earlier? Learning: What do I know now that I didn't at session start?

Key Principles

  • Specific over general - "No bulk AC add" not "CLI could be better"
  • Systemic over incidental - Focus on repeatable friction
  • Ask don't assume - User decides what's worth capturing
  • Brief on successes - Friction points are the value

Workflow Commands

bash
# Check current step
kspec workflow show

# Advance with notes
kspec workflow next --notes "..."

# Skip a step if not applicable
kspec workflow next --skip --notes "reason"

# Pause for later
kspec workflow pause

# Resume
kspec workflow resume

Integration

After reflection, observations can be:

  • Promoted to tasks: kspec meta promote @ref --title "..."
  • Resolved when addressed: kspec meta resolve @ref

Loop Mode

You are running in autonomous loop mode. Start the workflow:

bash
kspec workflow start @session-reflect-loop

Key Differences from Interactive Mode

  1. High confidence only - Only capture friction/successes you're certain about
  2. Search first - MUST search existing specs/tasks/inbox before capturing anything
  3. No user prompts - Skip discussion step, auto-resolve decisions
  4. Lower volume - Better to capture nothing than capture noise

Workflow Steps

  1. Review session - What worked well, what caused friction
  2. Search existing - For each potential capture:
    bash
    kspec search "<keyword>"
    
    If already tracked, skip it.
  3. Capture high-confidence items only
    • Clear friction pattern you encountered multiple times? Capture it
    • Uncertain or one-off issue? Skip it
    • Success pattern worth replicating? Capture it
  4. Exit - Don't wait for user confirmation

Exit Conditions

  • Session reviewed - Reflection complete (normal exit)
  • Nothing to capture - No high-confidence items identified
  • All already tracked - Search found existing coverage

What NOT to Capture

  • Vague observations ("could be better")
  • One-time issues that won't recur
  • Things you're unsure about
  • Anything already tracked in specs/tasks/inbox

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results