Auto Review Completed Todos

Topic: OpenCode Plugins · Tags: todo, review, productivity, automation

Auto-detect when all session todos are completed and send a review message. Fires once per session.

Site: dracon.uk

About

opencode-auto-review-completed-todos is an OpenCode plugin that automatically triggers a session review when all tasks are completed. It listens for OpenCode's internal todo.updated events — the same system used by the AI and UI — and sends a visible chat message prompting the AI to summarize the session.

Designed as the yin to opencode-todo-reminder's yang: where todo-reminder nudges when tasks remain incomplete, this plugin triggers a review when all tasks are done.

What it does

Listens for OpenCode's internal todo.updated events — whenever the todowrite tool creates, updates, or completes todos. When all todos have status completed or cancelled, it sends a review message to the chat. If new pending todos appear before the debounce fires, the review is cancelled.

This works with any todo source: the AI creating/checking todos via the todowrite tool, the user checking boxes in the UI, or the internal todo-reminder plugin.

Design philosophy: Pairs with opencode-todo-reminder as its complement. Todo-reminder nudges when tasks remain incomplete; this plugin triggers a review when all tasks are done.

Install

Option 1 — npm (recommended)

npm install -g opencode-auto-review-completed-todos

Register in opencode.json:

"plugin": [
  "opencode-auto-review-completed-todos@latest"
]

OpenCode will automatically fetch and cache the latest version from npm. No manual file copying needed.

Option 2 — manual

cp opencode-auto-review-completed-todos.js ~/.config/opencode/plugins/

Register in opencode.json (version will be pinned to whatever you copied):

"plugin": [
  "opencode-auto-review-completed-todos"
]

Restart OpenCode.

Configuration

{
  "plugin": [
    ["opencode-auto-review-completed-todos", {
      "debounceMs": 500,
      "maxSessions": 100,
      "maxRetries": 2
    }]
  ]
}

Option	Type	Default	Description
debounceMs	number	500	Wait after the last completed todo before sending message
maxSessions	number	100	Maximum concurrent sessions to track (oldest 20% evicted when limit is reached)
maxRetries	number	2	Retry prompt delivery this many times on failure (with backoff)

How it works

Architecture

Per-session state is minimal:

SessionState
├── reviewFired: boolean         ← prevent double-fire
├── debounceTimer: Timer | null  ← debounce before sending
└── createdAt: number            ← used for LRU eviction when maxSessions reached

Event handling

Event	Action
todo.updated	Check all todos. If all completed/cancelled → schedule message. If any pending → cancel scheduled message.
session.deleted / session.error / session.compacted	Clean up session state

Review trigger flow

1. User/AI completes todos via OpenCode's todowrite tool
2. todo.updated event fires with updated todo list
3. Plugin checks todos.every(t => t.status === "completed" || t.status === "cancelled")
4. If all done → debounce timer starts (default 500ms, configurable via debounceMs)
5. Timer fires → sends message to chat: "All tasks in this session have been completed. Please perform a final review..." (with retry on failure). The AI is explicitly instructed to create todos for any issues found, not just note them.
6. AI responds with a session review summary

How the message appears

Uses client.session.prompt() with synthetic: false — the same approach as opencode-todo-reminder. The message appears as a normal chat message that the AI responds to naturally.

Yin and yang

Plugin	When it triggers	What it does
opencode-todo-reminder	Todos remain incomplete	Nudges to complete tasks
opencode-auto-review-completed-todos	All todos complete	Triggers review

Status

STABLE — Published to npm (v1.0.5). Ready for production use.

Plugin:

• Detects todo.updated events from OpenCode's internal todowrite tool
• Sends visible chat message when all todos are completed
• AI responds with a session review summary
• Explicitly instructs AI to create todos for any issues found (not just note them)
• Debounces to avoid premature triggering
• Fires only once per session
• Configurable maxSessions cap prevents memory leak
• Retry with exponential backoff on prompt failure
• Structured logging for observability
• Cleans up on session.deleted, session.error, session.compacted

Files

Path	Description
~/.cache/opencode/packages/	OpenCode's npm package cache (used when referencing @latest)
~/.config/opencode/plugins/opencode-auto-review-completed-todos.js	Plugin file (manual install only)
~/.npm-global/lib/node_modules/opencode-auto-review-completed-todos/	npm package (global install)
~/Dev/opencode-auto-review-completed-todos/	Git-tracked source

Troubleshooting

Plugin firing twice (duplicate review prompts):

• This happens if the plugin is registered both as a local file in ~/.config/opencode/plugins/ and as an npm package reference. OpenCode loads both, causing double-fire.
• If using @latest in opencode.json, do not also copy the .js file to ~/.config/opencode/plugins/.
• To fix: remove the local file from ~/.config/opencode/plugins/ and clear OpenCode's cache at ~/.cache/opencode/packages/, then restart.

Plugin not loading:

• Verify opencode.json has "opencode-auto-review-completed-todos" (or "opencode-auto-review-completed-todos@latest") in the plugin array
• Ensure ~/.config/opencode/plugins/opencode-auto-review-completed-todos.js exists (manual install only)
• Only the .js file should be in the plugins directory — do not copy the .ts file

Message not appearing:

• Todos must be created via OpenCode's todowrite tool (not raw text like - [ ])
• All todos must have status "completed" or "cancelled"
• Plugin fires only once per session (new session needed after firing)

Requirements

• OpenCode with plugin support
• No additional npm dependencies

License

This project is dual-licensed:

• AGPL-3.0-only — See LICENSE for the full text. This is the default license for open source use.
• Commercial License — For organizations that prefer not to comply with AGPLv3's source disclosure requirements. See COMMERCIAL-LICENSE.md for details.

By contributing to this project, you agree to the terms in CLA.md.