648,523 mods indexed. 535,803 measured for token cost, 648,393 scanned.

Every mod for your coding agent, measured.

Every other catalogue counts. This one measures.

Skills, subagents, commands, hooks, Cursor rules and MCP servers for Claude Code, Cursor, Codex and Copilot. Each page says what the mod costs you in tokens on every session, what a security scan found and why, whether it is the original or one of the copies, and what changed since last week.

npx agentmods cost tells you what your current setup costs every session

Token cost per session Scanned, with reasons Original or copy Changes tracked
648,523 Mods indexed, across nine kinds and every major agent
146,197Repositories crawled, each re-checked and re-hashed
535,803Token cost counted locally for every mod that has a body
62,796Copies and forks identified, linked to their originals and demoted
Stars this week

Moved this week

The six mods whose repositories gained the most stars in the last seven days — movement, not accumulated fame.

K-Dense-AI/scientific-agent-skills

Skill Claude CodeCodex

Query and download public cancer imaging data from NCI Imaging Data Commons. Invoke for any question about IDC collections, cancer imaging datasets, DICOM data access, radiology (CT, MR, PET) or pathology AI training sets, metadata queries, visualization, or license checks — even when the user doesn't explicitly…

40k +1.4k yesterday A 75 tokens original MIT

apply

03

MadsLorentzen/ai-job-search

Command Claude Code

You are orchestrating a two-agent job application workflow. The job posting is provided below as $ARGUMENTS (either a URL or pasted text).

39k +967 today A 0 tokens original MIT

scrape

04

MadsLorentzen/ai-job-search

Skill Claude CodeCodex

Finds new job postings matching your profile via installed portal-search CLIs (LinkedIn, local job boards, and any skills added with /add-portal). Deduplicates across runs. Triggers on: job scrape, find jobs, search jobs, new jobs, job search, scrape jobs, /scrape.

39k +967 changed yesterday A 67 tokens original MIT

MadsLorentzen/ai-job-search

Agent Claude Code

Use this agent when the user needs to perform research tasks, gather information from external sources, or investigate topics that require web searches and synthesis of information.

39k +967 today A 37 tokens original MIT

MadsLorentzen/ai-job-search

Instructions file

Claude Code instructions for MadsLorentzen/ai-job-search, covering job application assistant for [yourname], role, candidate profile, identity and education.

39k +967 today A 2,161 tokens original MIT
Fewest tokens

Cheapest of what’s moving

Mods people are picking up this week, sorted by what they add to every session before you have asked anything. No other catalogue can offer this sort, because no other catalogue counts.

epic

01

gastownhall/beads

Command

Epic management commands.

27k +29 yesterday A 3 tokens original MIT

query

02

huifer/WellAlly-health

Command Claude Code

A command for searching a personal medical-record data store by type, date, recency, or abnormal results. It reads an index and related JSON files, then presents the matching records and summary counts.

935 +1 1mo ago A 4 tokens original MIT

dev-handlers

06

civitai/civitai

Command Claude Code

Give context for creating handlers.

7.2k +1 today A 5 tokens original Apache-2.0
The four measurements

What a listing does not tell you

Every catalogue shows that a mod exists and how many stars sit behind it. None of them tells you what it costs, whether it is safe, whether it is the original, or whether it changed under you.

What it costs, in tokens

Per session and once invoked, counted locally with o200k_base and priced per model. A stack of six agents is twenty thousand tokens before you have typed anything.

skills: 52 tokens average, preloaded
agents: 1 718, every session

A scan with its reasons

Hidden Unicode, instruction-override phrasing, curl piped into a shell, recursive deletes, credential paths. A grade from A to F, and the exact line behind every finding.

not a badge: the matched line
scan, not verified

Original, or one of the copies

Bodies are simhashed and clustered. The one with the most behind it is the original; the rest say how identical they are and point at it instead of competing with it.

indexed: everything
shown: the ones that differ

What changed since last week

Hashed on every crawl, with the history kept rather than the latest state alone. A supply-chain change to an agent config is a question of when, not whether.

every version: hash + date

Stars do not decide the order

Stars belong to the repository, not the mod: one repo with 170k stars and 139 skills would own the first three pages. Reputation enters through log10 and carries a quarter of the weight.

reputation 25 · freshness 20 · safety 20
originality 15 · quality 10 · traction 10

Nothing is hosted here

Every mod stays where its author put it and every page links to the source file. Bodies appear only under a permissive licence; no LICENSE means all rights reserved, and those pages say so.

we index. authors publish.
FAQ

Frequently
asked questions

What a mod is, what the numbers mean, and what this catalogue does and does not do.

Anything you add to a coding agent to change how it behaves: a skill, a subagent, a slash command, a hook, a Cursor rule, an instructions file, an MCP server, a plugin bundle. They are all the same shape of thing — a file you drop in that the agent then reads — so they belong in one catalogue rather than nine.

Because a mod is not free to have installed. An agent or a Cursor rule is injected into the context of every session whether or not you use it, and a stack of them can be twenty thousand tokens before you have asked anything. Every page here prints what the mod costs per session and what it costs once invoked, counted locally with the o200k_base tokenizer.

It is a static scan, not an audit, and the page says so. It looks for hidden Unicode, instruction-override phrasing, shell pipes into a downloaded script, recursive deletes, privilege escalation, credential paths and encoded payloads. Every finding is printed with the line that produced it, so you can judge whether it matters in context rather than trusting a letter.

Because most of them are. Bodies are simhashed and near-identical documents are clustered; the one with the most behind it is treated as the original and the rest point at it. A catalogue that lists nine copies of the same skill as nine results is making the search harder, not easier.

No. Every mod stays where its author put it, and every page links to the source file. Bodies are shown only where the repository carries a permissive licence; a repository with no LICENSE is all rights reserved, so those pages carry the metadata, the measurements and a link.

Mod your agent

Find the one you need, and know what it costs

648,523 mods, measured four ways. Search by name, or sort a whole category by what it adds to your context.