skill

Name: skill
Author: unbrowse-ai

730stars63forks

How to install skill?

unbrowse-ai/unbrowse/skill

$npx -y skills add unbrowse-ai/unbrowse --skill skill

Installs into the current project.

Summary

The action engine of the internet. Unbrowse is the open-source action layer for AI agents: it learns a site's internal API routes from real browsing, then replays them as fast, cheap, indexed routes (cache hit under 200ms) instead of re-driving a browser. Capture once, replay eve

Files · 1

View on GitHub

SKILL.md

1	# Unbrowse
2
3	Unbrowse is the action engine of the internet: the open-source action layer that turns
4	websites into reusable, indexed API routes for agents. Teach a route once by browsing,
5	store sanitized route metadata, replay it on later calls. A replay is about
6	30x faster and 90x cheaper than a fresh browser session (peer-reviewed: 3.6x mean speedup,
7	5.4x median over Playwright across 94 live domains, 18 domains under 100ms;
8	[Internal APIs Are All You Need](https://unbrowse.ai/whitepaper)).
9
10	## CLI quick paths (read `unbrowse --help` once, then stop)
11
12	The shipped CLI uses flat top-level commands. Do not prefix with `build` / `act` / `eval` / `act` — those legacy verb forms are not the primary surface.
13
14	\| You want \| Command \|
15	\|---\|---\|
16	\| One internet result (default) \| `unbrowse "task" --url <url>` or `unbrowse get "task" --url <url>` \|
17	\| A URL's contents \| `unbrowse fetch <url>` \|
18	\| Route/debug details \| `unbrowse resolve --intent "..." --url "..."` \|
19	\| Pick a specific endpoint \| `unbrowse resolve ... --no-execute` then `unbrowse execute --skill ID --endpoint ID` \|
20	\| Real DOM (forms, clicks) \| `unbrowse go <url>` → `snap` / `click` / `fill` / `submit` → `close` \|
21	\| First visit / miss \| `unbrowse capture --url <url> --intent "..."` \|
22	\| Login once \| `unbrowse auth <login_url>` \|
23	\| Bootstrap on install \| `unbrowse setup` \|
24	\| Health check \| `unbrowse health` \|
25
26	Browser-backed commands (`fetch`, `go`, `capture`, `auth`) need Chrome/Chromium installed. If `fetch` fails with a kuri/Chrome error, use `unbrowse get "task" --url <url>` (HTTP-first resolve path) or install Chrome and re-run `unbrowse setup`.
27
28	## The flow (load-bearing): ONE call by default. Resolve+execute for control. One capture on a miss.
29
30	For almost every read/search task ("find/get/list X on a site"), the FASTEST path is ONE
31	call. Let the runtime resolve the route, fill the holes, escalate if needed, and return the
32	structured result. Do NOT hand-run resolve, then fetch, then parse the page yourself.
33
34	unbrowse "<what you want>" --url "<site>" # bare natural-language: the one-hole front door
35	unbrowse get "<what you want>" --url "<site>" # identical, explicit form
36
37	Worked example, "homemade food on Carousell" (ONE call returns priced listings):
38
39	unbrowse "homemade food listings with prices and links" --url "https://www.carousell.sg/homemade-food/q/"
40
41	That single call runs resolve -> execute (or a direct fetch / one capture on a miss) and
42	returns the data. A real session that instead did `resolve` (8s, zero results on an
43	unindexed site) then hand-fetched and hand-parsed the page burned 1m41s for what one call
44	does. If you are writing a loop over URLs or piping fetch output through grep/python, stop:
45	you skipped the one-call path.
46
47	When you must PICK a specific endpoint (several routes, a mutation, explicit params), use the
48	two-call explicit path:
49
50	1. `unbrowse resolve --intent "<what you want>" --url "<site>" --no-execute` -> ranked shortlist.
51	2. `unbrowse execute --skill <id> --endpoint <id> [-p key=val ...]` -> replay it.
52
53	On a genuine MISS (no indexed route, a first visit, an anti-bot site), do ONE escalation:
54
55	unbrowse capture --url "<site>" --intent "<what you want>"
56
57	That drives the browser once and INDEXES the route. First visit to an uncached site pays a
58	capture tax (seconds); every visit after is a route-cache hit (<200ms). `resolve` on an
59	uncached site WILL miss (count 0) - that is expected; escalate with one capture, never a
60	fetch loop. The manual steps (`go`, `snap`, `click`, `sync`) exist, but prefer the single `capture`.
61
62	### STOP rules: this is exactly where agents waste minutes
63
64	- Do NOT `curl`, `WebFetch`, or `fetch` in a loop, or scrape pages by hand. Use the one-call
65	`unbrowse "task" --url`, or resolve + execute, or one capture. If you are writing a loop over
66	URLs or piping fetch output through grep/python, you are flailing: stop.
67	- Do NOT probe ports (`curl localhost:6969`), run `mcp serve`, or babysit a daemon. The CLI
68	runs in-process.

People who liked this also viewed

6 picks

Type

skill

730stars63forks

View on GitHub

How to install skill?

unbrowse-ai/unbrowse/skill

$npx -y skills add unbrowse-ai/unbrowse --skill skill

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Unbrowse
2
3	Unbrowse is the action engine of the internet: the open-source action layer that turns
4	websites into reusable, indexed API routes for agents. Teach a route once by browsing,
5	store sanitized route metadata, replay it on later calls. A replay is about
6	30x faster and 90x cheaper than a fresh browser session (peer-reviewed: 3.6x mean speedup,
7	5.4x median over Playwright across 94 live domains, 18 domains under 100ms;
8	[Internal APIs Are All You Need](https://unbrowse.ai/whitepaper)).
9
10	## CLI quick paths (read `unbrowse --help` once, then stop)
11
12	The shipped CLI uses flat top-level commands. Do not prefix with `build` / `act` / `eval` / `act` — those legacy verb forms are not the primary surface.
13
14	\| You want \| Command \|
15	\|---\|---\|
16	\| One internet result (default) \| `unbrowse "task" --url <url>` or `unbrowse get "task" --url <url>` \|
17	\| A URL's contents \| `unbrowse fetch <url>` \|
18	\| Route/debug details \| `unbrowse resolve --intent "..." --url "..."` \|
19	\| Pick a specific endpoint \| `unbrowse resolve ... --no-execute` then `unbrowse execute --skill ID --endpoint ID` \|
20	\| Real DOM (forms, clicks) \| `unbrowse go <url>` → `snap` / `click` / `fill` / `submit` → `close` \|
21	\| First visit / miss \| `unbrowse capture --url <url> --intent "..."` \|
22	\| Login once \| `unbrowse auth <login_url>` \|
23	\| Bootstrap on install \| `unbrowse setup` \|
24	\| Health check \| `unbrowse health` \|
25
26	Browser-backed commands (`fetch`, `go`, `capture`, `auth`) need Chrome/Chromium installed. If `fetch` fails with a kuri/Chrome error, use `unbrowse get "task" --url <url>` (HTTP-first resolve path) or install Chrome and re-run `unbrowse setup`.
27
28	## The flow (load-bearing): ONE call by default. Resolve+execute for control. One capture on a miss.
29
30	For almost every read/search task ("find/get/list X on a site"), the FASTEST path is ONE
31	call. Let the runtime resolve the route, fill the holes, escalate if needed, and return the
32	structured result. Do NOT hand-run resolve, then fetch, then parse the page yourself.
33
34	unbrowse "<what you want>" --url "<site>" # bare natural-language: the one-hole front door
35	unbrowse get "<what you want>" --url "<site>" # identical, explicit form
36
37	Worked example, "homemade food on Carousell" (ONE call returns priced listings):
38
39	unbrowse "homemade food listings with prices and links" --url "https://www.carousell.sg/homemade-food/q/"
40
41	That single call runs resolve -> execute (or a direct fetch / one capture on a miss) and
42	returns the data. A real session that instead did `resolve` (8s, zero results on an
43	unindexed site) then hand-fetched and hand-parsed the page burned 1m41s for what one call
44	does. If you are writing a loop over URLs or piping fetch output through grep/python, stop:
45	you skipped the one-call path.
46
47	When you must PICK a specific endpoint (several routes, a mutation, explicit params), use the
48	two-call explicit path:
49
50	1. `unbrowse resolve --intent "<what you want>" --url "<site>" --no-execute` -> ranked shortlist.
51	2. `unbrowse execute --skill <id> --endpoint <id> [-p key=val ...]` -> replay it.
52
53	On a genuine MISS (no indexed route, a first visit, an anti-bot site), do ONE escalation:
54
55	unbrowse capture --url "<site>" --intent "<what you want>"
56
57	That drives the browser once and INDEXES the route. First visit to an uncached site pays a
58	capture tax (seconds); every visit after is a route-cache hit (<200ms). `resolve` on an
59	uncached site WILL miss (count 0) - that is expected; escalate with one capture, never a
60	fetch loop. The manual steps (`go`, `snap`, `click`, `sync`) exist, but prefer the single `capture`.
61
62	### STOP rules: this is exactly where agents waste minutes
63
64	- Do NOT `curl`, `WebFetch`, or `fetch` in a loop, or scrape pages by hand. Use the one-call
65	`unbrowse "task" --url`, or resolve + execute, or one capture. If you are writing a loop over
66	URLs or piping fetch output through grep/python, you are flailing: stop.
67	- Do NOT probe ports (`curl localhost:6969`), run `mcp serve`, or babysit a daemon. The CLI
68	runs in-process.

People who liked this also viewed

6 picks

Type