Code quality
A code quality review your agent cannot skim
Ask a coding agent to review a codebase and it reads a sample, forms an impression and writes a confident summary. This package replaces the impression with a procedure: two assessments, several hundred individual checks, each one its own step that the engine hands to the agent and will not move past until the findings for that check are written down and reviewed.
The first assessment is ISO/IEC 5055:2021, the international standard that measures software quality at the source level. The second is the Clean Code guidelines from Robert C. Martin's book, chapter by chapter. Both end in a compiled report with a prioritized remediation plan, and every finding cites the file and line it came from.
Code Quality
Free
All plans · included with your license
Teams taking on an unfamiliar codebase, preparing for an audit, or wanting a baseline before a refactor. It is the free on-ramp package: it comes with the Starter tier, so it is also the simplest way to see what a workflow on rails looks like.
What runs
The workflows inside
ISO 5055 Code Quality Assessment
Runs the four CISQ measures — Maintainability, Performance Efficiency, Reliability and Security — each checking every quality measure element (a CWE) in its own step, then compiles an executive summary with per-measure totals and a prioritized remediation plan.
Clean Code Assessment
Twelve chapter workflows — Meaningful Names, Functions, Comments, Formatting, Objects and Data Structures, Error Handling, Boundaries, Unit Tests, Classes, Systems, Emergence, and Smells and Heuristics — each appending reviewed findings to a shared log, then a compile step that dedupes, ranks and verifies every recommendation.
How a run goes
Step by step, in the order the engine enforces
- 1
Point it at a directory
The agent calls ccw_start with the package's workflow and the path to review. The engine takes over the order of events from here.
- 2
One check, one step
The engine hands the agent a single weakness or chapter item at a time, with the instruction file for exactly that check. Findings are appended to a log on disk; the accumulated log never passes through a prompt, so a long review does not degrade as the context fills.
- 3
Review before proceeding
Each chapter's findings go through a review node that loops fix-and-re-review, up to three iterations, before the engine opens the next chapter.
- 4
Compile and deliver
A final workflow dedupes across chapters, ranks by severity and effort, verifies that each recommendation is actionable, and writes the report — gated by a last deliverable review.
What you get
At the end of a run
- ✓An ISO 5055 report with per-measure totals and a prioritized remediation plan
- ✓A Clean Code report with findings per chapter, every one citing file and line
- ✓The raw findings log, so the numbers can be audited rather than trusted
Read next
Related packages
- Architecture Basic →
A twenty-nine chapter Clean Architecture assessment of your codebase — SOLID, components, boundaries — every checklist item a verified step. Free from Pro.
- Security Review →
A source-level OWASP Top 10 assessment and a STRIDE threat model with trust boundaries, built by your coding agent one verified step at a time. Free from Pro.