How to use Claude to run reviewed coding workflows
A team uses Claude for planning and implementation, sol for adversarial review, and continuity documents for session management. It reports higher productivity, fewer bugs, and clearer documentation.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Plan [subfeature] against [specification] and [coding standards]. Review the plan's impact before implementation, then implement and test. Ask [independent reviewer] to assess coverage, quality, architecture, and documentation. Write a continuity document before ending the session.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- K0balt
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports approximately twice the productivity at about $300 per developer per month, with fewer bugs and clearer documentation; independent verification is not supplied.
- Human involvement
- The team specifies work, maintains review standards, manages session continuity, and intervenes when sessions go off track.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Specify the work and create an implementation plan.
- Review the plan against specifications and the impact of changes.
- Implement the change and run tests.
The author describes a development workflow centered on specification, repeated review, and session continuity. Work begins with a specification and a plan, followed by a review comparing the plan with requirements and the potential impact of changes. Claude handles planning and generation, while sol performs adversarial review.
After implementation and testing, the team reviews coverage, coding-guideline compliance, code smells, architectural boundaries, and documentation consistency. Only then does it plan the next subfeature. Review sessions receive the coding standards without prior session context, which the author considers an important part of the approach.
The team also maintains continuity documents for onboarding and wrapping up sessions. Before context compaction, it writes transition documents. When a session goes off track, the team sidelines it and later asks it to review committed work, analyze impact, and suggest remediation.
The author reports roughly twice the productivity at approximately $300 per developer per month, alongside fewer bugs and clearer documentation. These are the team's reported metrics rather than independently verified findings.
Who it's for: Teams prepared to invest in structured review and context management around AI-assisted development.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · K0balt
Frequently asked questions
How does the team divide work between models?
The author says Claude performs planning and generation, while sol handles adversarial review with coding standards and no prior session context.
What happens when a session goes off track?
The team sidelines it, waits for master to advance, then requests a review of committed work, impact analysis, and remediation.
More details
Remaining steps
- Review coverage, code quality, architectural compliance, and documentation.
- Write continuity and transition documents for subsequent sessions.
About this account
Claude plans and generates code while a separate model conducts repeated adversarial reviews within a structured development process.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=49161712. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: K0balt on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude · sol · Continuity documents · Coding standards
Adversarial review · Context management · Development workflow
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.