Agent Skills for competitive GTM work - battlecards, win-loss, positioning, pricing, launches, deal prep and sales enablement. Installs into Claude Code, Codex, Copilot CLI, Cursor, Gemini CLI and any Agent Skills client. Works standalone; uses live evidence when LMTY is connected.. It is open-source on GitHub.

Is skills free to use?

skills is open-source under the MIT license, so it is free to use.

What category does skills belong to?

skills is listed under skills in the Claudeers registry of Claude-compatible tools.

Platforms	cli, api, desktop, web
Operating systems	—
AI compatibility	claude
License	MIT
Pricing	open-source
Language	Python

Competitive Skills by LMTY

Use your AI assistant for competitive work you already do: researching competitors and market shifts, building positioning and messaging, reviewing pricing and launches, creating battlecards, preparing deals, analyzing wins and losses, and keeping sales enablement current.

Add the pack once, then ask in plain English, "Give me this week's competitive briefing." Your assistant selects the matching skill and follows a repeatable method for completing the work. The skills work in Claude Code, ChatGPT and Codex, GitHub Copilot, Cursor, Gemini CLI and any assistant that supports Agent Skills; the pack itself is an Agent Plugin.

Every skill works on its own. Connect LMTY and each one starts from your tracked competitors, kept current with what has changed and the evidence behind it, instead of a blank page.

What you can ask

Skill	Ask something like	You get
`competitive-market-briefing`	"Give me this week's competitive briefing."	A calm digest of what actually changed, and what to watch
`competitive-deep-dive`	"Get me fully up to speed on Northwind."	One competitor, understood and evidenced
`competitive-market-shift-analysis`	"Everyone is adding AI. Is this a real shift or just noise?"	A verdict: noise, one move, an emerging shift, or a new expectation
`competitive-executive-market-brief`	"Board meeting Friday. What market changes could blindside me?"	What leadership must decide, with the uncertainty shown
`competitive-positioning`	"Our competitors all sound like us now. Pressure-test our positioning."	A position tested against the alternatives buyers really consider
`competitive-messaging-audit`	"Does our homepage still differentiate us?"	What still works, what went generic, what buyers would say instead
`competitive-pricing-packaging-review`	"Northwind raised prices. What should we look at before changing ours?"	The market evidence a pricing decision needs, and the decision named
`competitive-launch-readiness`	"Here's the PRD. What do I need to know before we launch?"	Whether it's parity, catch-up or real differentiation, and how to frame it
`competitive-battlecard`	"Build a battlecard for us vs Northwind."	A card a rep can use with two minutes before a call
`competitive-objection-response`	"The buyer just said Northwind is cheaper. What do I say?"	The facts checked, a response, and the question to ask next
`competitive-deal-prep`	"Prep me for the Globex call tomorrow - we're up against Northwind."	One deal, prepared for the next meeting
`competitive-poc-rfp-plan`	"We're going into a POC against Northwind. What should we prove?"	An evaluation planned around the criteria that will decide it
`competitive-enablement-refresh`	"Northwind changed their pricing. What in our talk tracks is now wrong?"	The lines that changed, with proof and date
`competitive-asset-audit`	"Audit our competitive materials. Which claims are stale?"	Every competitive claim, marked and ranked by exposure
`competitive-win-loss-analysis`	"Why are we losing to Northwind?"	What buyers actually said, counted, with the exceptions kept
`competitive-roadmap-review`	"Competitors keep adding this. Does it matter for our roadmap?"	Product questions from market and customer evidence, not a copied feature list

You do not pick a skill; the right one is chosen for you. If a request needs more than one, the assistant does the part that fits and tells you which skill to run next.

You can also call a skill directly: /lmty:competitive-battlecard Northwind in Claude Code and GitHub Copilot, $competitive-battlecard in Codex, @ in ChatGPT.

Connect LMTY

Without LMTY, a skill researches the competitor from scratch each time and tells you the answer is a snapshot. With LMTY connected, it already knows your product, your tracked competitors, what changed and when, and the sources behind each claim.

Without LMTY	With LMTY
Asks who "we" are and which competitor	Already knows your product and your competitor set
Researches a snapshot, dated today	Reads tracked market state with a history of changes
"Not observed"	"Not observed, last checked on `<date>`"
Sources are whatever it found this session	Every finding links back to its public source, with when LMTY observed it.

LMTY tracks a product's competitors and keeps a current report plus a log of what moved.

The connection is already set up for you. This pack ships with the LMTY server configured, so in Claude Code, ChatGPT and Codex, GitHub Copilot CLI, Cursor and Gemini CLI your assistant will ask you to sign in, in your browser - ChatGPT and Codex as part of installing, the others the first time you open a session with the plugin enabled. Approve it once and every skill here uses it from then on.

No skill needs LMTY to run, and none will refuse work without it. If you never connect, the skills research from scratch and say so.

Connecting sends your requests to LMTY's server. LMTY's Privacy policy and Terms of service cover what happens to them.

Connecting by hand, or with an API token

If your assistant installed the skills through npx skills add, it did not see the server. Add it to your assistant's MCP configuration as a Streamable HTTP server at https://mcp.lmty.com/mcp; sign-in is discovered automatically.

If you would rather not use the browser sign-in, LMTY can issue a token from your account settings. Add the server to your own configuration with an Authorization: Bearer <token> header rather than relying on the bundled one.

Install

Each install below is the LMTY plugin: the full skill pack plus the LMTY connection. Start with the tool you already use.

ChatGPT and Codex

In the ChatGPT desktop app, the LMTY marketplace appears as a source in the Plugins Directory once it has been added; install the plugin from there and it is available in Codex and in ChatGPT Work.

To add the marketplace, from the Codex CLI:

codex plugin marketplace add LMTYdotcom/skills
codex plugin add lmty@lmty-plugins

Skills are invoked as $competitive-battlecard in Codex and with @ in ChatGPT.

Claude Code

Inside a Claude Code session, type /plugin marketplace add LMTYdotcom/skills, then /plugin install lmty@lmty-plugins.

Or run these two commands in your terminal:

claude plugin marketplace add LMTYdotcom/skills
claude plugin install lmty@lmty-plugins

To get a newer version later:

claude plugin marketplace update lmty-plugins
claude plugin update lmty@lmty-plugins

To try it without installing, download the release zip and run claude --plugin-dir lmty-<version>.zip.

More tools and the command line

GitHub Copilot, Cursor, Gemini CLI, and any other assistant

GitHub Copilot CLI

copilot plugin marketplace add LMTYdotcom/skills
copilot plugin install lmty@lmty-plugins

Or install straight from the repository, without a marketplace:

copilot plugin install LMTYdotcom/skills

Cursor

Cursor loads the pack as an Agent Plugin. Clone or unzip it into ~/.cursor/plugins/local/lmty, restart Cursor, and the skills appear under Customize. Teams and Enterprise workspaces can add the repository as a team marketplace instead, from Dashboard -> Plugins.

Gemini CLI

gemini extensions install https://github.com/LMTYdotcom/skills

OpenCode, Antigravity, Windsurf, Amp, Kiro, Cline and the rest

The skills CLI installs the skills into whichever assistants it finds on your machine:

npx skills add LMTYdotcom/skills

That asks which skills and which assistants you want. To take the whole pack into every assistant it finds without being asked, add --all, which is shorthand for --skill '*' --agent '*' -y:

npx skills add LMTYdotcom/skills --all

To install just one skill, add --skill and its name:

npx skills add LMTYdotcom/skills --skill competitive-battlecard

This route installs the skills only. To also connect LMTY, add the MCP server to your assistant by hand; see Connect LMTY.

Download

Every release on the releases page includes lmty-<version>.zip. Unzip it and point your assistant at the folder, or copy a single folder out of skills/ into wherever your assistant keeps skills (for example ~/.claude/skills/). Each skill is self-contained.

Releases are immutable and signed. To check that a zip is the one this repository published:

gh release verify-asset v1.0.0 lmty-1.0.0.zip -R LMTYdotcom/skills

Versions and updates

Most installs track the main branch, not a release, and every client decides for itself when to pull a change:

Client	Installs from	Updates when
Claude Code	`main`	you run `claude plugin update`, and the version has changed
ChatGPT and Codex	`main`	you refresh the marketplace
GitHub Copilot CLI	`main`	you run `copilot plugin update`
Cursor	the branch the marketplace tracks	that branch moves
Gemini CLI	the latest release	a newer release is published
skills CLI	`main`	you run `npx skills update`

main is always a releasable state - every push is validated - so tracking it is the intended default.

To pin to a release instead, name it at install time:

npx skills add LMTYdotcom/skills#v1.0.0
gemini extensions install https://github.com/LMTYdotcom/skills --ref=v1.0.0
codex plugin marketplace add LMTYdotcom/skills --ref v1.0.0

In the skills CLI, # takes a tag, branch or commit, while @ picks a skill, so LMTYdotcom/skills#v1.0.0@competitive-battlecard is one skill at one version. npx skills update then keeps re-checking the ref you pinned rather than moving you forward.

How the skills behave

Competitive work goes wrong in a few repeatable ways. Most of what is in these skills exists to stop one of them.

Absence is not evidence. A pricing page that does not mention audit logging says something about the page, not the product. The skills write "not observed in the sources checked, as of a date," never "does not have."
No invented motive. A competitor raised prices; why is unknowable without a statement from them, and the skills say so instead of guessing.
Dates travel with claims. Every material fact carries the date it was observed, and coverage is tracked separately from freshness.
Fact and interpretation are kept apart. Observed fact, interpretation, implication, and action are labeled, and any ranking is marked as the skill's own judgment.
Recommendations name the decision, not the answer. Without your internal numbers and deal evidence, the useful output is the question and the evidence needed to settle it.
Empty sections stay empty. "No validated objections yet - pull the last 20 calls where this competitor was named" is more useful than three plausible ones.
They hand off. A request that straddles two skills gets the half this skill owns and the name of the one that owns the rest.

How they were tested

Every skill was run against three models from different vendors, with a fourth model grading the output against written criteria, and then run live in Claude Code against the real LMTY server. Across the pack, the right skill is chosen for 95% of example requests, including the ones written to sit on the boundary between two neighbors.

The test criteria are kept out of this repository on purpose: a skill should not ship with its own answer key.

Contributing

Bug reports and suggestions are welcome as issues. If you want to change a skill, CONTRIBUTING.md explains how the repository is laid out and what a change needs before it merges. Taking part means following the Code of Conduct.

Need help rather than a change? SUPPORT.md says where to ask.

License

MIT. See LICENSE.

skills

Install with your AI

// compatibility