You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: .github/aw/serena-tool.md
+2-57Lines changed: 2 additions & 57 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -60,14 +60,7 @@ imports:
60
60
61
61
### 1. Activate Serena First
62
62
63
-
**Always call activate_project before other Serena tools:**
64
-
65
-
```javascript
66
-
// activate_project tool
67
-
{
68
-
"path": "/home/runner/work/gh-aw/gh-aw"
69
-
}
70
-
```
63
+
Always call `activate_project` before other Serena tools, passing the workspace path.
71
64
72
65
### 2. Combine with Other Tools
73
66
@@ -96,35 +89,6 @@ imports:
96
89
cache-memory: true # Store analysis history
97
90
```
98
91
99
-
## Common Patterns
100
-
101
-
### Pattern 1: Find All Function Usages
102
-
103
-
```
104
-
1. Use find_symbol to locate function definition
105
-
2. Use find_referencing_code_snippets to find call sites
106
-
3. Analyze patterns
107
-
```
108
-
109
-
### Pattern 2: Code Quality Analysis
110
-
111
-
```
112
-
1. Use get_symbols_overview on multiple files
113
-
2. Use find_symbol for similar function names
114
-
3. Use search_for_pattern for duplicate logic
115
-
4. Identify consolidation opportunities
116
-
```
117
-
118
-
### Pattern 3: Daily Code Analysis
119
-
120
-
```
121
-
1. Load previous state from cache-memory
122
-
2. Select files using round-robin or priority
123
-
3. Use Serena for semantic analysis
124
-
4. Save findings to cache
125
-
5. Generate improvement tasks
126
-
```
127
-
128
92
## Common Pitfalls
129
93
130
94
❌ **Using Serena for non-code files** - Use `edit` for YAML/JSON/Markdown
@@ -134,23 +98,4 @@ cache-memory: true # Store analysis history
134
98
135
99
## Supported Languages
136
100
137
-
Primary languages with full LSP features:
138
-
- `go` (gopls)
139
-
- `typescript` (TypeScript/JavaScript)
140
-
- `python` (jedi/pyright)
141
-
- `ruby` (solargraph)
142
-
- `rust` (rust-analyzer)
143
-
- `java`, `cpp`, `csharp`
144
-
145
-
See `.serena/project.yml` for complete list (25+ languages).
146
-
147
-
## Decision Tree
148
-
149
-
```
150
-
Task requires code semantics/structure?
151
-
├─ NO → Use bash/edit/view
152
-
└─ YES
153
-
├─ Simple text search/replace? → Use grep/bash
154
-
├─ Config/data files? → Use edit
155
-
└─ Symbol/structure/semantic patterns? → Use Serena ✅
156
-
```
101
+
Full LSP: `go`(gopls), `typescript`, `python` (jedi/pyright), `ruby` (solargraph), `rust` (rust-analyzer), `java`, `cpp`, `csharp`. See `.serena/project.yml` for full list (25+).
Copy file name to clipboardExpand all lines: .github/aw/subagents.md
+4-18Lines changed: 4 additions & 18 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -10,17 +10,7 @@ Inline sub-agents let you define specialised agents directly inside a workflow m
10
10
11
11
## Enabling the Feature
12
12
13
-
Inline sub-agent extraction and restoration steps are enabled by default:
14
-
15
-
```yaml
16
-
---
17
-
engine: copilot
18
-
---
19
-
```
20
-
21
-
> `features.inline-agents` is deprecated and should be removed. Inline sub-agent extraction is always on, so this field is unnecessary.
22
-
>
23
-
> `inline-sub-agents: false` is not supported and fails compilation. Remove the field.
13
+
Inline sub-agent extraction is always on — no config needed. The deprecated `features.inline-agents` field and the unsupported `inline-sub-agents: false` field both fail compilation; remove them.
24
14
25
15
---
26
16
@@ -87,12 +77,10 @@ The engine is detected at compile time from the `engine:` field and injected as
87
77
88
78
## MCP Access in Sub-Agents
89
79
90
-
Sub-agents **do not have their own MCP servers**. They run within the parent workflow's agent environment but without independent tool configuration.
91
-
92
-
For sub-agents to perform useful work they typically need access to the file system and shell. The following tools must be enabled on the parent workflow:
80
+
Sub-agents **do not have their own MCP servers** — they run within the parent's agent environment without independent tool config. To give them file system and shell access, enable on the parent workflow:
93
81
94
-
-**`cli-proxy: true`** — enables the GitHub CLI proxy so the sub-agent can make authenticated GitHub API calls via `gh`. Strongly recommended for any sub-agent that reads or writes repository content.
95
-
-**`tools.github.mode: gh-proxy`** — routes GitHub API calls through the gh proxy sidecar; required for the sub-agent to operate on private repositories or to use the GitHub MCP toolset.
82
+
-**`cli-proxy: true`** — GitHub CLI proxy for authenticated `gh` calls. Recommended for any sub-agent that reads/writes repo content.
83
+
-**`tools.github.mode: gh-proxy`** — routes GitHub API calls through the gh proxy sidecar; required for private repos or the GitHub MCP toolset.
96
84
97
85
```yaml
98
86
---
@@ -199,6 +187,4 @@ changes. Return a bulleted list, one bullet per file.
199
187
200
188
- Sub-agents do not support `engine:`, `tools:`, `network:`, or `mcp-servers:` fields — those are stripped at runtime.
201
189
- Sub-agents cannot define their own safe-output jobs.
202
-
-`features.inline-agents` is deprecated and should be removed; inline sub-agent upload/restore is always generated.
203
-
-`inline-sub-agents: false` is rejected at compile time; inline sub-agents cannot be disabled.
204
190
- Sub-agent blocks must appear in the main workflow file body; they are not resolved inside imported shared files.
0 commit comments