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
description: "Initialize a project with recommended rules and configuration",
14
-
content: `Please analyze this codebase and create an AGENTS.md file containing:
13
+
description: "Analyze codebase and create concise AGENTS.md files for AI assistants",
14
+
content: `<task>
15
+
Please analyze this codebase and create an AGENTS.md file containing:
15
16
1. Build/lint/test commands - especially for running a single test
16
17
2. Code style guidelines including imports, formatting, types, naming conventions, error handling, etc.
18
+
</task>
17
19
18
-
Usage notes:
19
-
- The file you create will be given to agentic coding agents (such as yourself) that operate in this repository. Make it about 20 lines long.
20
-
- If there's already an AGENTS.md, improve it.
21
-
- If there are Claude Code rules (in CLAUDE.md), Cursor rules (in .cursor/rules/ or .cursorrules), or Copilot rules (in .github/copilot-instructions.md), make sure to include them.
22
-
- Be sure to prefix the file with the following text:
20
+
<initialization>
21
+
<purpose>
22
+
Create (or update) a concise AGENTS.md file that enables immediate productivity for AI assistants.
23
+
Focus on project-specific, non-obvious information. Prioritize brevity and scannability.
24
+
25
+
Usage notes:
26
+
- The file you create will be given to agentic coding agents (such as yourself) that operate in this repository
27
+
- Keep the main AGENTS.md concise - aim for about 20 lines, but use more if the project complexity requires it
28
+
- If there's already an AGENTS.md, improve it
29
+
- If there are Claude Code rules (in CLAUDE.md), Cursor rules (in .cursor/rules/ or .cursorrules), or Copilot rules (in .github/copilot-instructions.md), make sure to include them
30
+
- Be sure to prefix the file with: "# AGENTS.md\\n\\nThis file provides guidance to agents when working with code in this repository."
31
+
</purpose>
32
+
33
+
<todo_list_creation>
34
+
If the update_todo_list tool is available, create a todo list with these focused analysis steps:
35
+
36
+
1. Quick scan for existing docs
37
+
- AI assistant rules (.cursorrules, CLAUDE.md, AGENTS.md, .roorules)
38
+
- README and key documentation
39
+
40
+
2. Identify stack
41
+
- Language, framework, build tools
42
+
- Package manager and dependencies
43
+
44
+
3. Extract commands
45
+
- Build, test, lint, run
46
+
- Critical directory-specific commands
47
+
48
+
4. Map core architecture
49
+
- Main components and flow
50
+
- Key entry points
51
+
52
+
5. Document critical patterns
53
+
- Project-specific utilities
54
+
- Non-standard approaches
55
+
56
+
6. Extract code style
57
+
- From config files only
58
+
- Key conventions
59
+
60
+
7. Testing specifics
61
+
- Framework and run commands
62
+
- Directory requirements
63
+
64
+
8. Compile concise AGENTS.md
65
+
- Essential sections only
66
+
- Brief, scannable format
67
+
- Project-specific focus
68
+
69
+
9. Create mode-specific rule directories
70
+
- Create directory structures for the four core modes: .roo/rules-code/, .roo/rules-ask/, .roo/rules-architect/, .roo/rules-debug/
71
+
- Create mode-specific AGENTS.md files with rules specific to that mode's purpose and capabilities
72
+
- These rules should provide additive context and not just repeat the mode definitions
73
+
- Only include rules that you have high confidence are accurate, valuable, and non-obvious
74
+
75
+
Note: If update_todo_list is not available, proceed with the analysis workflow directly without creating a todo list.
76
+
</todo_list_creation>
77
+
</initialization>
23
78
24
-
# AGENTS.md
79
+
<analysis_workflow>
80
+
Follow the comprehensive analysis workflow to:
81
+
82
+
1. **Discovery Phase**: Find existing documentation and AI assistant rules
83
+
2. **Project Identification**: Identify language, stack, and build system
84
+
3. **Command Extraction**: Extract and verify essential commands
85
+
4. **Architecture Mapping**: Create visual flow diagrams of core processes
86
+
5. **Component Analysis**: Document key components and their interactions
87
+
6. **Pattern Analysis**: Identify project-specific patterns and conventions
88
+
7. **Code Style Extraction**: Extract formatting and naming conventions
89
+
8. **Security & Performance**: Document critical patterns if relevant
90
+
9. **Testing Discovery**: Understand testing setup and practices
91
+
10. **Example Extraction**: Find real examples from the codebase
92
+
</analysis_workflow>
25
93
26
-
This file provides guidance to agents when working with code in this repository.
94
+
<output_structure>
95
+
<main_file>
96
+
Create AGENTS.md with:
97
+
- Header: "# AGENTS.md\\n\\nThis file provides guidance to agents when working with code in this repository."
- New features must work within VSCode extension + webview architecture
201
+
- Provider implementations must be stateless and cacheable
202
+
- UI state management uses React hooks, not external state libraries
203
+
- Database schema changes require migrations in packages/evals/src/db/migrations/
204
+
- New packages must follow the existing monorepo structure in packages/
205
+
- API changes must maintain backward compatibility with existing provider contracts
206
+
\`\`\`
207
+
</mode_specific_files>
208
+
</output_structure>
27
209
28
-
Additionally, please:
29
-
1. **Create mode-specific rule directories** - Create directory structures for the four core modes: \`.roo/rules-code/\`, \`.roo/rules-ask/\`, \`.roo/rules-architect/\`, and \`.roo/rules-debug/\`
30
-
2. **Create mode-specific AGENTS.md files** - Within each of these four mode directories, research and then create an AGENTS.md file with rules specific to that mode's purpose and capabilities. These rules should provide additive context and not just repeat the mode definitions. Only include rules that you have high confidence are accurate, valuable, and non-obvious.
210
+
<quality_criteria>
211
+
- Include visual flow diagrams (ASCII/markdown) for architecture
212
+
- Provide actionable, step-by-step guides
213
+
- Focus on non-obvious, project-specific information
214
+
- Include real code examples from the project
215
+
- Be concise and scannable
216
+
- Adapt to the specific project needs
217
+
- Document only what's essential for productivity
218
+
</quality_criteria>
31
219
32
-
**For the complete list of available modes with detailed descriptions, refer to the system prompt.** The system prompt contains comprehensive information about each mode's purpose, when to use it, and its specific capabilities.
33
-
34
-
Example structure with specific instructions:
35
-
36
-
\\\`\\\`\\\`
37
-
AGENTS.md # General project guidance
38
-
.roo/
39
-
├── rules-code/
40
-
│ └── AGENTS.md # Code mode specific instructions
41
-
├── rules-debug/
42
-
│ └── AGENTS.md # Debug mode specific instructions
43
-
├── rules-ask/
44
-
│ └── AGENTS.md # Ask mode specific instructions
45
-
└── rules-architect/
46
-
└── AGENTS.md # Architect mode specific instructions
- New features must work within VSCode extension + webview architecture
92
-
- Provider implementations must be stateless and cacheable
93
-
- UI state management uses React hooks, not external state libraries
94
-
- Database schema changes require migrations in packages/evals/src/db/migrations/
95
-
- New packages must follow the existing monorepo structure in packages/
96
-
- API changes must maintain backward compatibility with existing provider contracts
97
-
\\\`\\\`\\\`
98
-
99
-
This structure provides both general project guidance and specialized instructions for the four core modes' specific domains and workflows.
100
-
`,
220
+
Remember: The goal is to create documentation that enables AI assistants to be immediately productive in this codebase, focusing on project-specific knowledge that isn't obvious from the code structure alone.`,
0 commit comments