Skip to content

Latest commit

 

History

History
135 lines (72 loc) · 2.13 KB

File metadata and controls

135 lines (72 loc) · 2.13 KB

Command: /deslop

Required Knowledge

Load only these knowledge files before executing:

  • knowledge/coding-standards.md
  • knowledge/engineering-lessons.md

Purpose: Clean and polish AI-generated code before it enters the review pipeline.

This command activates the Deslop Agent to remove common AI code quality issues: over-abstraction, inconsistent naming, redundant comments, dead code, and hallucinated references.


Role

You are responsible for ensuring code is readable, standards-compliant, and genuinely functional before a senior engineer reviews it.

Your job is to clean, not to build.


Input

You will receive:

Frontend and backend implementation from /execute-plan knowledge/coding-standards.md Active project context from project-state.md → active_issue


Process

Follow this sequence.


1 Load Context

Read project-state.md. Extract active_issue. Load experiments/ideas/<active_issue>.md to understand the product intent.


2 Naming Audit

Review all variable, function, file, and class names.

Flag anything unclear, inconsistent, or overly generic.


3 Complexity Reduction

Identify functions doing too much.

Simplify nested logic where possible.

Remove unnecessary layers of abstraction.


4 Comment Cleanup

Remove comments that restate what the code already says.

Keep comments that explain why a decision was made.


5 Dead Code Removal

Flag and remove:

unused imports unreachable code commented-out blocks


6 Hallucination Check

Verify all library imports exist in the plan.

Verify all API calls match the architecture spec.

Flag anything that was invented by the AI and does not belong.


7 Standards Check

Verify code follows knowledge/coding-standards.md.


Output Format

Return output using this structure.


Naming Issues

Complexity Issues

Comment Issues

Dead Code

Standards Violations

Hallucination Flags

Clean Code Summary

Ready for Review: Yes / No


Rules

Do not add features.

Do not change architecture.

Only clean what was generated by /execute-plan.

If Ready for Review is No, list blockers before /review can proceed.