Run a codebase-wide audit with Claude Code dynamic workflows
Added ⧉ Copyable prompt below
Our verdict
The right tool when a task is too big for one conversation: an auth sweep across every route, a large migration, a bug hunt. The plan lives in a script, so the run is resumable and repeatable, and findings are checked by other agents before you see them. It uses much more of your plan than a normal session, so start with one scoped directory.
- Who it's for
- Claude Code users on paid plans with large or legacy codebases
- Leverage
- Dozens of parallel agents on one task, with built-in cross-checking
- Effort
- Medium
- Date added
What it does
A dynamic workflow is a JavaScript script Claude writes that orchestrates many subagents at once; a runtime executes it in the background while your session stays responsive. Unlike subagents or skills, the script holds the loop and intermediate results, so it scales to dozens or hundreds of agents and can be saved as your own slash command. Claude Code also bundles /deep-research as a ready-made workflow.
How builders use it
- Try the bundled workflow first: /deep-research <question>, then /workflows to watch progress.
- Start a one-off workflow by putting the keyword ultracode in your prompt, scoped to one directory.
- Approve the plan when Claude Code asks, then drill into phases and agents from /workflows.
- If the run did what you wanted, press s in the progress view to save it as a command and rerun it later.
- On Pro, enable workflows from the Dynamic workflows row in /config.
What to expect
A single report of findings that other agents tried to refute, plus a saved script you can rerun after the next big change.
Copy this: Scoped audit
ultracode: audit every API endpoint under src/routes/ for missing auth checks and unvalidated input.
For each finding, have an independent agent try to refute it before it goes in the report.
Report file:line, the risk, and a suggested fix. Do not change any code.Related discoveries
- ToolCoding & Shipping Agents
Cursor Bugbot for PR review
An AI reviewer that flags bugs and security issues on every pull request.
Effort: Low
- ToolCoding & Shipping Agents
CodeRabbit
Independent AI code review, PR triage and change explanations.
Effort: Low
- SetupMulti-Step Agent Setups
Specialised subagents in Claude Code
Delegate review, testing or research to focused subagents with their own context.
Effort: Low⧉ Prompt
- ToolMulti-Step Agent Setups
LangGraph
Low-level framework for stateful, controllable agent workflows.
Effort: High
Want more like this?
Practical discoveries, without the noise.