Reflect — Mid-Conversation Reassessment
Reflect is a mid-conversation reflection skill that pauses execution and zooms out from detail-mode to honestly reassess direction, assumptions, and bias. It produces a flowing, conversational analysis covering macro perspective, gap analysis, reflective inquiry, bias check, and contextual alignment, ending with a clear recommendation to continue, pivot, or pause.
✨ What it does
- Pauses the current task before reassessing.
- Runs a 5-dimension analysis: macro perspective, gap analysis, reflective inquiry, bias check, and contextual alignment.
- Detects five cognitive biases (confirmation, sunk cost, anchoring, complexity, recency) with specific evidence.
- Produces flowing prose without headers or bullet lists.
- Ends with a mandatory directional recommendation: continue, pivot, or pause for a specific question.
- Optionally asks one clarifying question only when context is too thin to reassess.
🎯 When to use it
- When the user explicitly asks to reflect, step back, zoom out, or sanity check the conversation.
- When the conversation has gone deep into implementation details for many turns without a strategic check-in.
- When the user shows signs of being stuck, frustrated, or repeatedly hitting dead-ends.
- When you suspect the conversation may have drifted from the original goal or is missing something important.
🚀 How to use
Trigger the skill by saying 'reflect', 'step back', 'zoom out', 'are we missing something', 'sanity check this', or similar phrases. It can also be offered when the conversation has been deep in details for 10+ turns or the user seems stuck. The skill runs automatically on the full conversation history; no additional input is needed unless context is thin, in which case it asks one question about what to reassess.
Example prompts:
Reflect on where we are with this project.
Step back — are we on track?
📄 Output: A conversational reassessment with a clear recommendation to continue, pivot, or pause.
📦 Add this skill to Claude Code
# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills
# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/productivity/reflect/skills/reflect .claude/skills/reflectSkill source: productivity/reflect/skills/reflect/SKILL.md
⚠️ Good to know
Requires sufficient conversation context to reassess; if context is too thin, it will ask one clarifying question. It does not auto-invoke on implicit triggers—it offers to step back instead.
❓ FAQ
What does the skill do when invoked?
It pauses the current task and produces a flowing analysis covering macro perspective, gaps, reflective inquiry, bias check, and contextual alignment, ending with a recommendation to continue, pivot, or pause.
Does it always ask questions before reflecting?
No, it is low-intake and runs immediately when there is enough context. It only asks one clarifying question if the invocation context is too thin to reassess.
What if the current direction is actually fine?
The skill will state that clearly with specific reasoning and recommend continuing; it avoids manufacturing problems.
🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.