agent-browser

Name: agent-browser
Author: avibebuilder

109stars17forks

How to install agent-browser?

avibebuilder/claude-prime/agent-browser

$npx -y skills add avibebuilder/claude-prime --skill agent-browser

Installs into the current project.

Summary

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a websit

Files · 1

View on GitHub

SKILL.md

1	# Browser Automation with agent-browser
2
3	Use this skill to drive websites through the `agent-browser` CLI. Keep the main loop tight: inspect the page, act with refs, verify the result, and only pull deeper docs when the task actually needs them.
4
5	## Core workflow
6
7	Prefer `agent-browser` directly for speed. Use `npx agent-browser` only if it is not installed globally.
8
9	For most tasks, follow this loop:
10
11	1. Open the page
12	2. Wait for the relevant state
13	3. Snapshot with refs
14	4. Interact using those refs
15	5. Re-snapshot after page or DOM changes
16	6. Verify the outcome
17	7. Close the session when done
18
19	```bash
20	agent-browser open https://example.com/form
21	agent-browser wait --load networkidle
22	agent-browser snapshot -i
23	# Output: @e1 [input type="email"], @e2 [input type="password"], @e3 [button] "Submit"
24
25	agent-browser fill @e1 "user@example.com"
26	agent-browser fill @e2 "password123"
27	agent-browser click @e3
28	agent-browser wait --load networkidle
29	agent-browser snapshot -i
30	```
31
32	Chain commands with `&&` only when you do not need to inspect intermediate output. Good: `open && wait && screenshot`. Bad: `snapshot && click` when you still need to read the refs from the snapshot.
33
34	## Route fast
35
36	Read only the reference that matches the task:
37
38	\| Need \| Read \|
39	\| --- \| --- \|
40	\| Full command or flag lookup \| [references/commands.md](references/commands.md) \|
41	\| Ref lifecycle, stale refs, snapshot strategy \| [references/snapshot-refs.md](references/snapshot-refs.md) \|
42	\| Login flows, OAuth, 2FA, saved auth state \| [references/authentication.md](references/authentication.md) \|
43	\| Parallel sessions, state reuse, cleanup \| [references/session-management.md](references/session-management.md) \|
44	\| Recording, profiling, local files, config, iOS, security \| [references/advanced-usage.md](references/advanced-usage.md) \|
45	\| Proxy setup \| [references/proxy-support.md](references/proxy-support.md) \|
46	\| Recording workflows \| [references/video-recording.md](references/video-recording.md) \|
47	\| Profiling workflows \| [references/profiling.md](references/profiling.md) \|
48
49	## Golden path commands
50
51	Use these first; go to the command reference only when you need something more specific.
52
53	```bash
54	agent-browser open <url>
55	agent-browser wait --load networkidle
56	agent-browser snapshot -i
57	agent-browser click @e1
58	agent-browser fill @e2 "text"
59	agent-browser select @e3 "option"
60	agent-browser get url
61	agent-browser get text @e1
62	agent-browser diff snapshot
63	agent-browser screenshot --annotate
64	agent-browser close
65	```
66
67	## Refs are the default interaction model
68
69	The main value of `agent-browser` is that snapshots produce compact refs like `@e1`, `@e2`, `@e3`. Those refs are cheaper and more reliable than repeatedly reasoning from raw HTML or long selectors.
70
71	Treat refs as short-lived. Re-snapshot after anything that can change the page state, especially:
72	- navigation
73	- form submission
74	- opening dropdowns or modals
75	- lazy-loaded or client-rendered content
76
77	If a ref fails or the page looks different from what you expected, your next move is usually `agent-browser snapshot -i`, not another blind click.
78
79	For the full lifecycle and troubleshooting rules, read [references/snapshot-refs.md](references/snapshot-refs.md).
80
81	## Choose the lightest tool that still proves the result
82
83	Default order:
84	1. `snapshot -i` for structure and interactive targets
85	2. `get text`, `get url`, or `get title` for precise verification
86	3. `diff snapshot` when you need to confirm something changed
87	4. `screenshot --annotate` when layout, icon-only controls, canvas, or visual context matters
88	5. semantic locators or `eval` only when refs are unavailable or the task truly needs them
89
90	If you need semantic locators, JavaScript evaluation, local file access, annotated screenshots, or config details, jump to [references/advanced-usage.md](references/advanced-usage.md) and [references/commands.md](references/commands.md).
91
92	## Authentication: decide sensitivity first
93
94	Before filling any credential, classify the auth flow.
95
96	- Non-sensitive: localhost, staging, test accounts, or credentials the user explicitly provided for this task. The agent can usually fill these directly.
97	- Sensitive: production domains, real user accounts, OAuth/SSO, or anything where the agent should not handle the secret. In that case, reach the auth step, switch to a headed browser if needed, and let the user complete sign-in manually.
98
99	After either path su

People who liked this also viewed

6 picks

Type

agent-browser

109stars17forks

View on GitHub

How to install agent-browser?

avibebuilder/claude-prime/agent-browser

$npx -y skills add avibebuilder/claude-prime --skill agent-browser

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Browser Automation with agent-browser
2
3	Use this skill to drive websites through the `agent-browser` CLI. Keep the main loop tight: inspect the page, act with refs, verify the result, and only pull deeper docs when the task actually needs them.
4
5	## Core workflow
6
7	Prefer `agent-browser` directly for speed. Use `npx agent-browser` only if it is not installed globally.
8
9	For most tasks, follow this loop:
10
11	1. Open the page
12	2. Wait for the relevant state
13	3. Snapshot with refs
14	4. Interact using those refs
15	5. Re-snapshot after page or DOM changes
16	6. Verify the outcome
17	7. Close the session when done
18
19	```bash
20	agent-browser open https://example.com/form
21	agent-browser wait --load networkidle
22	agent-browser snapshot -i
23	# Output: @e1 [input type="email"], @e2 [input type="password"], @e3 [button] "Submit"
24
25	agent-browser fill @e1 "user@example.com"
26	agent-browser fill @e2 "password123"
27	agent-browser click @e3
28	agent-browser wait --load networkidle
29	agent-browser snapshot -i
30	```
31
32	Chain commands with `&&` only when you do not need to inspect intermediate output. Good: `open && wait && screenshot`. Bad: `snapshot && click` when you still need to read the refs from the snapshot.
33
34	## Route fast
35
36	Read only the reference that matches the task:
37
38	\| Need \| Read \|
39	\| --- \| --- \|
40	\| Full command or flag lookup \| [references/commands.md](references/commands.md) \|
41	\| Ref lifecycle, stale refs, snapshot strategy \| [references/snapshot-refs.md](references/snapshot-refs.md) \|
42	\| Login flows, OAuth, 2FA, saved auth state \| [references/authentication.md](references/authentication.md) \|
43	\| Parallel sessions, state reuse, cleanup \| [references/session-management.md](references/session-management.md) \|
44	\| Recording, profiling, local files, config, iOS, security \| [references/advanced-usage.md](references/advanced-usage.md) \|
45	\| Proxy setup \| [references/proxy-support.md](references/proxy-support.md) \|
46	\| Recording workflows \| [references/video-recording.md](references/video-recording.md) \|
47	\| Profiling workflows \| [references/profiling.md](references/profiling.md) \|
48
49	## Golden path commands
50
51	Use these first; go to the command reference only when you need something more specific.
52
53	```bash
54	agent-browser open <url>
55	agent-browser wait --load networkidle
56	agent-browser snapshot -i
57	agent-browser click @e1
58	agent-browser fill @e2 "text"
59	agent-browser select @e3 "option"
60	agent-browser get url
61	agent-browser get text @e1
62	agent-browser diff snapshot
63	agent-browser screenshot --annotate
64	agent-browser close
65	```
66
67	## Refs are the default interaction model
68
69	The main value of `agent-browser` is that snapshots produce compact refs like `@e1`, `@e2`, `@e3`. Those refs are cheaper and more reliable than repeatedly reasoning from raw HTML or long selectors.
70
71	Treat refs as short-lived. Re-snapshot after anything that can change the page state, especially:
72	- navigation
73	- form submission
74	- opening dropdowns or modals
75	- lazy-loaded or client-rendered content
76
77	If a ref fails or the page looks different from what you expected, your next move is usually `agent-browser snapshot -i`, not another blind click.
78
79	For the full lifecycle and troubleshooting rules, read [references/snapshot-refs.md](references/snapshot-refs.md).
80
81	## Choose the lightest tool that still proves the result
82
83	Default order:
84	1. `snapshot -i` for structure and interactive targets
85	2. `get text`, `get url`, or `get title` for precise verification
86	3. `diff snapshot` when you need to confirm something changed
87	4. `screenshot --annotate` when layout, icon-only controls, canvas, or visual context matters
88	5. semantic locators or `eval` only when refs are unavailable or the task truly needs them
89
90	If you need semantic locators, JavaScript evaluation, local file access, annotated screenshots, or config details, jump to [references/advanced-usage.md](references/advanced-usage.md) and [references/commands.md](references/commands.md).
91
92	## Authentication: decide sensitivity first
93
94	Before filling any credential, classify the auth flow.
95
96	- Non-sensitive: localhost, staging, test accounts, or credentials the user explicitly provided for this task. The agent can usually fill these directly.
97	- Sensitive: production domains, real user accounts, OAuth/SSO, or anything where the agent should not handle the secret. In that case, reach the auth step, switch to a headed browser if needed, and let the user complete sign-in manually.
98
99	After either path su

People who liked this also viewed

6 picks

Type