gm-oc 2.0.371 → 2.0.372

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/package.json +1 -1
  2. package/skills/browser/SKILL.md +17 -65
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "gm-oc",
3
- "version": "2.0.371",
3
+ "version": "2.0.372",
4
4
  "description": "State machine agent with hooks, skills, and automated git enforcement",
5
5
  "author": "AnEntrypoint",
6
6
  "license": "MIT",
package/skills/browser/SKILL.md CHANGED
@@ -1,36 +1,12 @@
1
1
  ---
2
2
  name: browser
3
- description: Browser automation via playwriter. Use when user needs to interact with websites, navigate pages, fill forms, click buttons, take screenshots, extract data, test web apps, or automate any browser task.
4
- allowed-tools: Bash(browser:*), Bash(exec:browser*)
3
+ description: Browser automation. Use when user needs to interact with websites, navigate pages, fill forms, click buttons, take screenshots, extract data, test web apps, or automate any browser task.
4
+ allowed-tools: Bash(exec:browser*)
5
5
  ---
6
6
 
7
- # Browser Automation with playwriter
7
+ # Browser Automation
8
8
 
9
- ## Two Pathways
10
-
11
- **Session management** — use `browser:` prefix via Bash for session lifecycle only.
12
-
13
- Create a session first:
14
-
15
- ```
16
- browser:
17
- playwriter session new --direct
18
- ```
19
-
20
- Returns a numeric session ID (e.g. `1`). Use that ID for all subsequent `exec:browser` calls.
21
-
22
- If `--direct` fails, the user needs Chrome running with debugging enabled:
23
- - Open `chrome://inspect/#remote-debugging` in Chrome, OR
24
- - Launch Chrome with `chrome --remote-debugging-port=9222`
25
-
26
- List active sessions:
27
-
28
- ```
29
- browser:
30
- playwriter session list
31
- ```
32
-
33
- **JS eval** — use `exec:browser` via Bash for ALL JavaScript execution. Never use `playwriter -s <id> -e '...'` for JS code — single-quote quoting fails on Windows CMD. The exec runner writes code to a temp file, avoiding all shell quoting issues.
9
+ Use `exec:browser` via Bash for all browser automation. The runtime provides `page`, `snapshot`, `screenshotWithAccessibilityLabels`, and `state` as globals. Sessions persist across calls automatically.
34
10
 
35
11
  ```
36
12
  exec:browser
@@ -38,25 +14,9 @@ await page.goto('https://example.com')
38
14
  await snapshot({ page })
39
15
  ```
40
16
 
41
- ```
42
- exec:browser
43
- const title = await page.title()
44
- console.log(title)
45
- ```
46
-
47
- State persists across `exec:browser` calls within a session. Never add shell quoting or escaping to the exec body — write plain JavaScript directly.
48
-
49
17
  ## Core Workflow
50
18
 
51
- Every browser automation follows this pattern:
52
-
53
- 1. **Create session**: `browser:\nplaywriter session new --direct` (note the returned ID)
54
- 2. **All JS code**: use `exec:browser` with plain JS body — navigate, interact, snapshot, extract
55
-
56
- ```
57
- browser:
58
- playwriter session new --direct
59
- ```
19
+ Navigate, snapshot to understand the page, then interact:
60
20
 
61
21
  ```
62
22
  exec:browser
@@ -70,14 +30,6 @@ await snapshot({ page })
70
30
 
71
31
  ## Common Patterns
72
32
 
73
- ### Navigation and Snapshot
74
-
75
- ```
76
- exec:browser
77
- await page.goto('https://example.com')
78
- await snapshot({ page })
79
- ```
80
-
81
33
  ### Screenshot with Accessibility Labels
82
34
 
83
35
  ```
@@ -98,12 +50,14 @@ console.log(JSON.stringify(items))
98
50
 
99
51
  ```
100
52
  exec:browser
101
- state.loginDone = false
102
- await page.goto('https://app.example.com/login')
103
- await page.fill('[name=user]', 'admin')
104
- await page.fill('[name=pass]', 'secret')
105
- await page.click('[type=submit]')
106
- state.loginDone = true
53
+ state.count = 0
54
+ await page.goto('https://example.com')
55
+ state.title = await page.title()
56
+ ```
57
+
58
+ ```
59
+ exec:browser
60
+ console.log(state.title, state.count)
107
61
  ```
108
62
 
109
63
  ### Console Monitoring
@@ -120,12 +74,10 @@ exec:browser
120
74
  console.log(JSON.stringify(state.consoleMsgs))
121
75
  ```
122
76
 
123
- ## Key Patterns for Agents
124
-
125
- **Always use `exec:browser`** for any JavaScript — never `playwriter -s <id> -e '...'` for JS code.
77
+ ## Key Rules
126
78
 
127
- **`browser:` prefix** is only for session management: `playwriter session new`, `playwriter session list`.
79
+ **Only `exec:browser`** — never run any browser CLI tool directly via Bash.
128
80
 
129
- **Session IDs are numeric**: `playwriter session new` returns `1`, `2`, etc. Use the exact returned value.
81
+ **Snapshot before interacting** — always call `await snapshot({ page })` to understand current page state before clicking or filling.
130
82
 
131
- **Snapshot before interacting**: always call `await snapshot({ page })` to understand current page state before clicking or filling.
83
+ **State persists** — `state` object and page session carry across multiple `exec:browser` calls.