ai-dev-worklog
v0.5.0
Published
Local worklog CLI for AI-assisted development sessions.
Maintainers
Readme
AI Dev Worklog
A local CLI that turns git evidence into a readable AI development worklog.
AI Dev Worklog scans the current git repository and writes a Markdown report, a JSON report, and a continuation prompt. It is useful when an AI-assisted coding session needs a local, reviewable record of what changed, what the task goal was, and how to continue.
The CLI is intentionally small and local-first: it reads git evidence, writes files under .ai-dev-worklog/, and keeps the output on your machine.
What It Does Now
ai-dev-worklog scan records:
- optional declared intent from
--intent - changed files from
git status --porcelain - tracked file changes from
git diff --name-status - diff summary from
git diff --stat - raw command output for the scanned git commands
- structured JSON fields for evidence, observed changes, validation status, risk signals, and next steps
- a continuation prompt file for the next development session
Quick Start
Run once with npx:
npx ai-dev-worklog scanAdd a task goal for the next session:
npx ai-dev-worklog scan --intent "Prepare the repository for an open-source release"Install globally:
npm install -g ai-dev-worklog
ai-dev-worklog scan --intent "Review local changes before opening a pull request"Run from source:
git clone https://github.com/alexan780/ai-dev-worklog.git
cd ai-dev-worklog
npm ci
npm run build
node dist/index.js scan --intent "Verify the local CLI from source"Local Output
Running ai-dev-worklog scan writes:
.ai-dev-worklog/
latest.md
latest.json
continue-prompt.mdlatest.md is the human-readable report.
latest.json is the machine-readable scan result. It uses a structured worklog schema that separates declared intent, git evidence, observed changes, validation status, risk signals, and next steps. The formal JSON Schema is available at schema/worklog.schema.json.
continue-prompt.md is a prompt-oriented summary for continuing the development session.
Example output is available in examples/basic-worklog.
Example summary:
## Summary
- Git status entries: 2
- Git diff name-status entries: 2
- Changed files: 2
- Working tree has changes: yes
- Validation: not_recorded
- Risk signals: 0
- Declared intent: Prepare the repository for an open-source releaseWhat The CLI Scans
The local CLI reads these git commands:
git status --porcelain
git diff --stat
git diff --name-statusThe scan uses local git evidence as the source of truth for observed code changes.
The CLI does not read chat windows, call AI APIs, upload code, or integrate with Codex App, Claude Code, Cursor, or MCP tools.
Privacy And Data Flow
AI Dev Worklog runs locally in the git repository where you call it.
- It runs local git commands.
- It writes local files under
.ai-dev-worklog/. - It does not upload source code or generated worklogs.
- It does not require an API key.
Why This Exists
When context breaks, tools switch, or a task becomes long, it can be hard to answer:
- What files actually changed?
- What git evidence is available?
- What still needs checking?
- How should the next session continue?
AI Dev Worklog turns AI-assisted coding sessions into local, reviewable development records.
Local Development
Install dependencies:
npm ciBuild the TypeScript CLI:
npm run buildRun tests:
npm testRun a local scan:
node dist/index.js scan --intent "Validate local development output"Generated build output goes to dist/.
Generated worklog output goes to .ai-dev-worklog/.
GitHub Actions runs the same install, build, and test checks on pushes to main and on pull requests.
What It Records
The CLI records observed local git evidence.
Current output includes:
- optional declared intent from the CLI
- changed files from
git status --porcelain - tracked file changes from
git diff --name-status - diff summary from
git diff --stat - raw command output for the scanned git commands
- a JSON worklog schema with declared intent, evidence, observed changes, validation status, risk signals, and next steps
- a continuation prompt file with changed files, command results, risk signals, remaining checks, and suggested next steps
Future versions may add command history and validation evidence.
What It Is
- Local first
- Evidence before summary
- Human-readable Markdown
- Machine-readable JSON
- Minimal dependencies
Roadmap
Released
v0.1 - Local Core
- Scan the current git repository
- Record changed files from local git evidence
- Generate
latest.mdandlatest.json - Keep all data local
v0.2 - Worklog Schema
- Define a stable JSON structure for AI coding worklogs
- Separate declared intent from observed evidence
- Add unified observed file changes
- Add validation status and basic risk signals
v0.3 - Continuation Prompt
- Generate a continuation prompt for the next development session
- Keep git evidence as the source of truth for observed code changes
v0.4 - Open-source Readiness
- Add GitHub Actions CI
- Add a formal JSON Schema file for
latest.json - Add a changelog and release history
- Improve README demo flow and roadmap clarity
v0.5 - Declared Intent
- Add
--intentfor recording the task goal in local worklog output - Include declared intent in Markdown and continuation prompt output
- Prepare npm metadata and public installation paths
Future Work
- Command history and validation evidence
- Optional workflow guides for AI coding tools
Status
This project is in early development. The CLI scans a git repository and generates a local Markdown worklog, a structured JSON scan result, and a continuation prompt.
The next milestones focus on strengthening the local CLI contract before adding optional workflow integrations.
