Agent Skills
Teach your AI coding agent to use LGTM.lol — fetch images, approve pull requests with an LGTM image, and submit new images. Works with Claude Code, Cursor, Codex CLI, and any agent supporting the Agent Skills standard. No API key required.
Install
npx skills add dongri/lgtm-skills
Or install a single skill:
npx skills add dongri/lgtm-skills --skill lgtm
Source: github.com/dongri/lgtm-skills
Available Skills
| Skill | Description |
|---|---|
lgtm |
Fetch a random LGTM image as Markdown, URL, or HTML — with format (gif / png / jpg / webp) and period filters |
lgtm-pr-approve |
Approve or comment on a GitHub/GitLab pull request with an LGTM image via the gh / glab CLI |
lgtm-submit |
Submit a new image to LGTM.lol by URL, or upload a local file (PNG / JPEG / GIF / WebP, up to 10 MB) |
Example Prompts
| What you say to your agent | Skill used |
|---|---|
| Get me an LGTM image | lgtm |
| Give me an animated LGTM gif from this month | lgtm |
| Approve PR #123 with an LGTM image | lgtm-pr-approve |
| Review this pull request and LGTM it if it looks good | lgtm-pr-approve |
| Upload ./cat.gif to lgtm.lol | lgtm-submit |
| Register https://example.com/funny.gif on lgtm.lol | lgtm-submit |
Skills vs. MCP
LGTM.lol also ships an MCP server. The two are complementary:
- Skills need zero setup — the agent calls the public HTTP API directly and learns complete workflows, like approving a pull request with an image.
- MCP exposes
get_random_lgtm_imageas a native tool for clients without shell access, such as the Claude web app.