retrospective

Retrospective

Run a structured retrospective to capture learnings and write them to memory for compound improvement over time.

Usage

/retrospective [scope]

scope (optional): what to retrospect on — e.g. "last sprint", "this feature", "this week", "the auth refactor". Defaults to the current project/conversation context.

Steps

1. Gather context — Read recent git history, task completions, and any existing memory files relevant to the scope. Ask the user for any additional context not visible in files.

2. What went well — Identify decisions, approaches, or habits that produced good outcomes. Be specific: not "communication was good" but "daily standups caught the API contract mismatch early."

3. What didn't go well — Identify friction, mistakes, or missed opportunities. Focus on patterns, not blame.

4. What we'd do differently — Concrete, actionable changes for next time. Each item should be specific enough to act on.

5. Key decisions made — Capture any significant decisions and why they were made. These are often lost otherwise.

6. Write to memory — For each significant learning, write a memory file in the project's memory/ directory using the standard frontmatter format:

   • feedback_*.md for process/approach learnings
   • project_*.md for project-specific context and decisions
   • Update MEMORY.md index with any new files

7. Summary — Present a concise summary of what was captured and where it was saved.

Memory Format

---
name: <short_name>
description: <one-line description of the learning>
type: feedback | project
---

<The learning itself — lead with the rule or fact>

**Why:** <what caused this learning or decision>
**How to apply:** <when this should influence future behavior>