Agent skill
free-claim-audit
Audit any "free", "open-source", "free forever" or "$0" claim about a tool, repo, model or website before adopting it, recommending it, or repeating the claim in your own content.
Filed under Content and SEO.
From hrshitkunwar-tech/mindlyft-gtm-skills · 3 skill entries · 0 · pushed 2026-10-22
What it does when it runs
Audit any "free", "open-source", "free forever" or "$0" claim about a tool, repo, model or website before adopting it, recommending it, or repeating the claim in your own content. A ten-failure-mode checklist plus a one-command GitHub repo audit (identity and renames, licence red flags, README cost and requirement lines, same-name copy check). Use when a post, video, tweet or colleague says a tool is free; when building a list of free tools; or when someone asks "is X actually free, open source, or OK to use commercially".
Automated analysis of the skill and the 1 file bundled beside it. A skill’s own description is written to be selected by an agent, so it describes the job and not the dependencies.
- Keys and connectors you must supply
- GITHUB_TOKEN
- Hosts it reaches
- api.github.com
- Tool permissions it declares
- No
allowed-toolsin the frontmatter. It does act, so it runs under whatever permissions your session already grants. - Actions present in the files
- shell
Ask about free-claim-audit
Opens your assistant with this page's recorded context. Check the linked sources before relying on it.
Install it
View source on GitHub ↗git clone --depth 1 --filter=blob:none --sparse https://github.com/hrshitkunwar-tech/mindlyft-gtm-skills.git /tmp/mindlyft-gtm-skills git -C /tmp/mindlyft-gtm-skills sparse-checkout set "skills/free-claim-audit" mkdir -p ~/.claude/skills/free-claim-audit cp -R "/tmp/mindlyft-gtm-skills/skills/free-claim-audit/." ~/.claude/skills/free-claim-audit/
Picked up without a restart. A project skill of the same name is shadowed by your personal one. For one repository only, swap ~/.claude/skills for .claude/skills. Claude Code docs ↗
The folder is the same in every client that implements the format — 46 of them — so if yours is not above, only the destination changes.
Before you install: this skill will not complete its job on a bare agent. It needs GITHUB_TOKEN, which you have to obtain separately.
The skill
Source on GitHub ↗Reproduced in full from hrshitkunwar-tech/mindlyft-gtm-skills/blob/7602db388dbc50055a937da66a3b98f398f0747a/skills/free-claim-audit/SKILL.md, which is licensed MIT (repository). 1,087 words, 6 headings.
Free Claim Audit
Social posts about "free" tools are marketing. The repository's own files, the licence text and the vendor's pricing page are the source of truth. A large share of "free" claims fail on inspection, usually in one of the ten ways below.
The ten ways "free" fails
| # | Failure | What it looks like | How to catch it |
|---|---|---|---|
| 1 | The repo does not exist, or it is the wrong one | A name shown on screen returns 404 | Query the GitHub API for repos/OWNER/NAME |
| 2 | A low-star copy was filmed instead of the original | Same project name, a handful of stars, while the original has thousands | Search the repo name sorted by stars; check the "forked from" field |
| 3 | The licence is not free for commercial use | Creative Commons NonCommercial, or a custom licence the API reports as NOASSERTION. AGPL-3.0 is fine to run internally but has obligations if you modify and offer it over a network | Read the LICENSE file itself, never only the API's SPDX field |
| 4 | The licence carves out territory, scale or use | Some model licences exclude whole regions or require a separate licence above a monthly-active-user threshold | Search the LICENSE for territory, "monthly active", "non-commercial", "research purposes" |
| 5 | The code is free, the weights or data are not | MIT code alongside model weights under a vendor licence, often gated behind a sign-up | Read the README licence section and the model card separately |
| 6 | Free to clone, paid to run | Requires a paid model plan, a paid scraping API, usage credits or a hosted service to do anything useful | README lines mentioning paid, pricing, credits, "requires", API keys; any model or vendor name visible in a demo |
| 7 | The name is withheld ("comment a keyword for the link") | Repo names blurred or cropped | Resolve by vendor org, description and star count; label the result "inferred", never "confirmed" |
| 8 | The "free tool" is the author's own email-gated funnel | Several separately branded single-page tools that all share one owner, ask for marketing consent and end by selling one paid product | Visit each domain; read the footer, the consent checkbox and the after-submit call to action |
| 9 | A public framework renamed as a proprietary system | "The N-step system" that turns out to be a well-known loop (OODA, PDCA, AIDA, Jobs-to-be-Done) with a filler step added, sold behind a consultation | Count the steps against known frameworks; a padding step like "repeat" is the tell |
| 10 | Provenance claims the project itself contradicts, or no pricing at all | A news piece credits a project to reverse-engineering while the README says it was a clean-room implementation; a product announced as "now free" with no pricing page | Search the README for clean-room, provenance, "not affiliated"; request /pricing and record the HTTP status |
Procedure
- Get the exact owner and name from the screen or a file, never from audio or a transcript. Speech-to-text mangles names. If the name is withheld, use check 7 and label the result as inferred.
- Run the repo audit (requires
curlandpython3; no token needed):
It prints identity (following renames), stars, the API licence, archived flag, last push date and fork parent; LICENSE lines matching common restriction terms; README lines about cost and requirements; and the top three same-name repositories by stars. Unauthenticated GitHub search is rate limited to a few requests a minute. If the output saysbash scripts/fca.sh OWNER/NAMESEARCH UNAVAILABLE, wait a minute and rerun; do not read it as "no copies exist". SetGITHUB_TOKENin your environment to raise the limit. - Read what it flags. The script finds lines; a human judges them.
NOASSERTIONalways means "open the LICENSE file and read it". - Price the runtime. For every "requires" line, open the vendor's own pricing page and write down the free-tier limit and the date you checked. For models that need a GPU, note the minimum VRAM.
- If the claim names websites rather than repos, visit each one. Record the owner (footer, terms page), what the sign-up consents you to, and where the after-submit flow sends you. A single-page tool that costs an email address plus marketing consent and ends by selling one product is failure mode 8. Describe it as "free tier, owned by X, leads to Y", not as a neutral find.
- Check the terms-of-service surface. Anything that automates a social network, scrapes behind a login, or sends email at volume carries account-restriction risk. Name that risk explicitly in the verdict.
- Write a one-line verdict:
REAL: free, permissive licence, no runtime bill.MIXED: <which failure modes>: usable, with the named caveats.HYPE: <what the files say instead>: the claim is contradicted by the source.
Pre-flight before shipping a verdict or any copy that says "free"
- The owner and name came from a screen or a file, not from audio.
- The audit ran, and its key lines are quoted with the date.
- The LICENSE text was read whenever the SPDX field is
NOASSERTIONor a custom licence. - Every runtime dependency has a price or "free tier: X, checked on <date>".
- Every website named as a "free tool" was visited, and its owner and paid destination are written down.
- The copy never says "free forever". Prefer precise wording such as "MIT licensed, free to self-host; runs on your own API key".
- AGPL, non-commercial and territory-restricted licences are named wherever the tool is recommended.
Where results go
- A list or catalogue of free tools: each entry carries its verdict, its licence and the date checked.
- A contradicted claim in something you saved or were sent: add a short correction note next to it so the claim does not get repeated later.
- A "free" claim in your own content: fix the copy before it ships.
Notes
- This is not legal advice. For anything you will redistribute, host for others or build a business on, have someone qualified read the licence.
- Licences and pricing change. A verdict is only as good as the date on it.
Files bundled with it
These load only when the skill asks for them, so they cost nothing until it runs.
Other skills for the same job
Different authors, same problem. Matched on the words in the skill name, across every library in the catalogue except this one.
- free-tools by coreyhaines31 · 57,918
- seo-audit by coreyhaines31 · 57,918
- ad-extension-audit by irinabuht12-oss · 6,620
- ai-visibility-audit by irinabuht12-oss · 6,620
- google-ads-audit by irinabuht12-oss · 6,620
- landing-page-audit by irinabuht12-oss · 6,620
- landing-page-audit-39 by irinabuht12-oss · 6,620
- linkedin-ads-audit by irinabuht12-oss · 6,620
Need help setting it up?
This page tells you what free-claim-audit does and what it needs. Cheetah builds the agent setup it runs inside: data, CRM, sequencing and the guardrails.
Book a call →The directory stays free. There is nothing gated behind this.