- Home
- Skills
- Code Quality & Review
- CloneSieve: Helper-Clone & Dead-Export Auditor
Works with the AI Agents you already use
CloneSieve: Helper-Clone & Dead-Export Auditor
CloneSieve: Helper-Clone & Dead-Export Auditor
Free
CloneSieve: Helper-Clone & Dead-Export Auditor
Example session with this skill installed
Zero-dependency 14-rule codebase structural function clone (Type-2 normalized + 4-gram Jaccard shingle), duplicate helper name, and zero-caller dead-export auditor + Claude Code and Codex blind-write transcript scanner. Audited across 3,300 files (4,544 functions) and 153 agent transcripts: catches 117 structural clone groups (DataWordmark 7x), 370 duplicate helper names, 39.0% blind function additions before Grep/Glob search, and 163 dead orphan exports while filtering 2,370 false alerts.
- Read your context and instructions
- Compiled the clonesieve
Zero-dependency 14-rule codebase structural function clone (Type-2 normalized + 4-gram Jaccard shingle), duplicate helper name, and zero-caller dead-export auditor + Claude Code and Codex blind-write transcript scanner. Audited across 3,300 files (4,544 functions) and 153 agent transcripts: catches 117 structural clone groups (DataWordmark 7x), 370 duplicate helper names, 39.0% blind function additions before Grep/Glob search, and 163 dead orphan exports while filtering 2,370 false alerts.
Connects securely to your tools. The creator never sees your data.
About this skill
Codebases managed by AI agents often suffer from "helper sprawl," where agents re-implement the same utility functions repeatedly because they fail to search the repository first. This tool identifies structural duplicates and dead exports to reclaim maintainability and reduce token waste.
What it does
- Structural Clone Detection finds identical function skeletons across multiple files using Type-2 normalization.
- Near-Duplicate Analysis identifies similar logic using 4-gram Jaccard shingle matching with an 80% threshold.
- Dead Export Auditing flags zero-caller orphan exports and symbols that are exported but only used internally.
- Transcript Forensics scans agent session logs to identify "blind writes" where functions were created before any repository search occurred.
- Registry Generation builds a canonical helper index table for use in agent instruction files like
CLAUDE.md.
How it works
- Run a repository scan to identify structural clones, name collisions, and orphaned code.
- Analyze agent transcripts to verify if duplicates were caused by a lack of initial repo searching.
- Consolidate identified clones into shared modules and prune dead exports based on the audit report.
- Generate and embed a Canonical Helper Registry to prevent agents from creating future duplicates.
Frameworks & tools
Built for TypeScript, JavaScript, React, and Python environments. It integrates with terminal-based agent transcripts and provides JSON output for CI/CD pipelines.
Why this beats prompting it yourself
Manual prompts often miss structural similarities that aren't exact string matches. This tool uses normalized token hashing and Jaccard similarity to find clones that use different variable names or formatting, which agents typically overlook.
Use cases
- Auditing a repository after heavy AI-assisted development to prune redundant utility functions.
- Enforcing code quality in CI by failing builds when new structural clones are introduced.
- Generating a source-of-truth helper index to improve agent context and reduce hallucinated utilities.
Known limitations
Excludes functions shorter than 25 normalized tokens to avoid flagging boilerplate getters. It also skips standard Next.js and SST file conventions to prevent false positives on routing exports.
Skips functions under 25 tokens. Ignores Next.js/SST routing conventions and barrel exports. Requires session transcripts for habit analysis.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 9 days ago
- Free with an Agensi account
Requires
Permissions
Zero-dependency 14-rule codebase structural function clone (Type-2 normalized + 4-gram Jaccard shingle), duplicate helper name, and zero-caller dead-export auditor + Claude Code and Codex blind-write transcript scanner. Audited across 3,300 files (4,544 functions) and 153 agent transcripts: catches 117 structural clone groups (DataWordmark 7x), 370 duplicate helper names, 39.0% blind function additions before Grep/Glob search, and 163 dead orphan exports while filtering 2,370 false alerts.