Dedupe cluster

Skill aisflows/skill-cleaner/skills/dedupe-cluster

Clean messy AI-agent skill folders into a usable SKILL.md library.

Install

npx -y skills add aisflows/skill-cleaner --skill dedupe-cluster

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use as step 3 of AIS FLOWS Skill Library Cleaner to find exact duplicates, near-duplicates, old variants, runtime copies, and overlapping skill groups, producing DEDUP_GROUPS.md and DEDUP_GROUPS.json.

SKILL.md

1.9 KB, 360 tokens by cl100k_base, as published. Nobody here has run it

Skill Library Cleaner / Dedupe & Cluster

Role

Group the messy pile into useful decision clusters.

Input:

machine/INVENTORY.json
machine/RISK_BLOCKLIST.json

Output:

reports/DEDUP_GROUPS.md
machine/DEDUP_GROUPS.json

Cluster Types

Detect:

exact duplicate folder or file;
same name in SKILL.md;
same behavior with different folder names;
runtime copies of one skill;
old draft vs newer variant;
broad overlapping skills;
source dump copies;
private/local variants.

Keep exact duplicates separate from overlaps. Do not recommend physical archive for an overlap cluster unless the evidence shows the members are safe replacements for each other.

Required Group Fields

Each group must have:

group id;
group label;
member paths;
likely canonical candidate;
archive candidates;
blocked members;
reason;
confidence;
next action.

Decision Routes

Allowed routes:

archive-exact-duplicate
merge-near-duplicate
keep-separate
manual-dedupe-review
split-active-vs-blocked
reference-only
backlog

Use manual-dedupe-review when names or categories overlap but responsibilities differ. Use archive-exact-duplicate only when the same behavior/content is duplicated with high confidence.

Confidence Rules

high: same name/content/purpose and one clear canonical candidate.
medium: strong overlap but possible role differences.
low: weak keyword similarity only.

Low-confidence groups must not produce archive operations.

Pass Condition

The cleaner knows what should become one canonical item and what should leave the active shelf.