Your First ADW Run

Run one issue manually before setting up automation. This proves the system works and shows you what each phase produces.

Step 1 — Create a GitHub Issue

Your repo on github.com → Issues → New Issue:

Title: Add row count to query results

Body:
When a SQL query returns results, display the total row count
above the results table. Example: "42 rows returned"

Note the issue number (e.g. #1).

Step 2 — Run the ADW

bash

cd adws
uv run adw_plan_build.py 1

What happens:

A unique 8-char ADW ID is generated (e.g. a3f9k2m1)
Issue is fetched and classified as /feature
Spec written to specs/issue-1-adw-a3f9k2m1-sdlc_planner-row-count.md
Implementor reads the spec and builds the feature
Commits pushed and PR opened on GitHub

Step 3 — Review the PR

GitHub → Pull Requests. The PR has:

Branch: feat-1-a3f9k2m1-add-row-count-to-query-results
Spec file committed alongside the code
Commit messages linked to the issue

Read the spec. Check the code. Merge when satisfied.

What Just Happened

The ADW ID (a3f9k2m1) ties everything together:

agents/
└── a3f9k2m1/
    ├── adw_state.json            # Persistent state file
    ├── a3f9k2m1_plan_spec.md     # Implementation plan
    ├── planner/
    │   └── raw_output.jsonl      # Claude Code session transcript
    └── implementor/
        └── raw_output.jsonl      # Implementation session transcript

The state file enables resuming, retrying, or chaining phases independently:

bash

# Resume a failed build phase
uv run adw_build.py 1 a3f9k2m1

# Add testing after the fact
uv run adw_test.py 1 a3f9k2m1

# Add review after the fact
uv run adw_review.py 1 a3f9k2m1

ADW Tracking

Each run gets a unique 8-character ID (e.g., a1b2c3d4) that appears in:

Issue comments: a1b2c3d4_ops: Starting ADW workflow
Output files: agents/a1b2c3d4/sdlc_planner/raw_output.jsonl
Git commits and PRs
Branch name: feat-456-e5f6g7h8-add-user-authentication

Your First ADW Run ​

Step 1 — Create a GitHub Issue ​

Step 2 — Run the ADW ​

Step 3 — Review the PR ​