godmode-self-reflect

name	godmode:self-reflect
description	Session reflection — review what was accomplished this session, surface patterns and surprises, and write a structured retrospective. Use when asked to "reflect", "self-reflect", "what did we do this session", or at natural session close points.
requires	[]
next	["pattern-learner","mistake-tracker"]

Self-Reflect

Review the current session's work and produce a structured retrospective written to .ctx/godmode/reports/reflect/reflect-<YYYY-MM-DD>.md.

When to Use

• At the end of a working session before handoff
• When asked to "reflect" or "what did we do today"
• After a significant batch of commits lands (e.g. post-wave integration)
• Before cutting a release to sanity-check what actually changed

Step 1: Collect session evidence

Gather the raw material for the reflection:

# Commits since session start (last 24h is a safe proxy; adjust if session spans days)
git log --oneline --since="24 hours ago"

# Files changed across those commits
git diff --stat HEAD~1 HEAD

Use the Glob tool to list any files in the working scratch dir:

Glob: pattern=".ctx/_WORKING_DIR/*"

Current task graph state — use godmode task list:

godmode task list

Outstanding blocked tasks:

godmode task list --json | rg blocked

Also read .ctx/HANDOFF.*.yaml if present — it captures intent from session start.

Step 2: Identify what was accomplished

From the commit log and diff stat, extract:

• Shipped: concrete changes that landed (commits, features, fixes, TODOs added)
• Unfinished: tasks still running or pending, files modified but not committed
• Discarded / pivoted: things started but abandoned (check git stash list)

Step 3: Surface patterns and surprises

Answer these questions honestly based on the evidence:

1. What took longer than expected? — look for repeated edits to the same file, or multiple fix-up commits on the same topic.
2. What went smoothly? — first-pass successes, clean test runs, no hook failures.
3. What was discovered that wasn't in the original plan? — TODOs added, unexpected coupling found, agents that surfaced issues not on the task graph.
4. What would speed up the next session? — missing tools, unclear specs, slow feedback loops, friction points.

Step 4: Write the retrospective

Write to .ctx/godmode/reports/reflect/reflect-<YYYY-MM-DD>.md (overwrite if exists for today):

# Session Reflection — <YYYY-MM-DD HH:MM>

## Shipped

- <bullet per commit or logical change group>

## Unfinished

- <pending tasks, uncommitted changes, stashed work>
  - None if graph is clean and worktree is clean.

## Patterns & Surprises

### Took longer than expected

- <finding or "nothing notable">

### Went smoothly

- <finding>

### Discovered mid-session

- <unexpected finding, new TODO, coupling, etc.>

### Next session speedups

- <concrete suggestion — a missing test, a flaky tool, an unclear spec>

## Task graph snapshot

<paste output of `godmode task list`>

## Open questions

- <anything unresolved that the next session should address first>

Use the Write tool to create the file. Print a short summary (Shipped / Unfinished / Top insight) to stdout as well.

After writing the report, update the report index. Read .ctx/godmode/reports/godmode-reports.index.json, add the filename to categories.reflect.files (if not already present), and write it back with the Write tool.

Step 5: Update task graph (if applicable)

If any tasks were completed during the session but not marked done:

godmode task done <id> --commit <sha> --notes "<brief>"

If new work was discovered that belongs on the graph:

godmode task add "<title>" --crate-name <crate>

Do not invent tasks to fill the graph — only add what was genuinely discovered.

Guardrails

• Do not modify source code during reflection — read-only except for the .ctx/ report and task graph state updates.
• Do not rewrite history or amend commits to make the session look cleaner than it was.
• If the session had no commits, say so — an empty "Shipped" section is honest and useful.
• The reflection is for the next session's benefit, not performance review. Be precise, not flattering.
• Never write (fill in next session) or any deferred placeholder. Every section must contain actual content from this session. If a section genuinely has nothing to report, write - Nothing notable. — not a deferral. Deferred placeholders make the reflection useless to the next agent.