|
1
|
--- |
|
2
|
name: skill-authoring |
|
3
|
description: Create reusable skills that extend agent capabilities. Use when the user asks to create a skill, teach you something reusable, or save instructions for future tasks. |
|
4
|
--- |
|
5
|
|
|
6
|
# Skill Authoring |
|
7
|
|
|
8
|
Create skills to save reusable knowledge, procedures, and workflows that persist across sessions. Skills are instructions for future versions of yourself—write them as if teaching a fresh instance how to handle a task. |
|
9
|
|
|
10
|
## When to Use |
|
11
|
|
|
12
|
- User asks to "create a skill" or "teach you how to do X" |
|
13
|
- User wants to save instructions for repeated future use |
|
14
|
- A workflow should be reusable across sessions |
|
15
|
|
|
16
|
## How to Create a Skill |
|
17
|
|
|
18
|
1. Choose a skill name (lowercase, hyphens only, e.g., `code-review`) |
|
19
|
2. Write the SKILL.md file to `.agents/skills/SKILLNAME/SKILL.md` |
|
20
|
|
|
21
|
## SKILL.md Format |
|
22
|
|
|
23
|
```markdown |
|
24
|
--- |
|
25
|
name: skill-name |
|
26
|
description: What this skill does. When to use it. |
|
27
|
--- |
|
28
|
|
|
29
|
# Skill Title |
|
30
|
|
|
31
|
Instructions, examples, and workflows go here. |
|
32
|
``` |
|
33
|
|
|
34
|
### Frontmatter Requirements |
|
35
|
|
|
36
|
- `name`: Max 64 chars, lowercase letters/numbers/hyphens only |
|
37
|
- `description`: Max 1024 chars, include WHAT it does and WHEN to use it |
|
38
|
|
|
39
|
## Writing Tips |
|
40
|
|
|
41
|
- **Be concise**: Only include info you don't already know |
|
42
|
- **Match specificity to fragility**: Exact commands for fragile ops, general guidance for flexible tasks |
|
43
|
- **Include examples**: Input/output pairs help with output quality |
|
44
|
- **Keep under 500 lines**: For longer skills, reference separate files |
|
45
|
|
|
46
|
## Skill Locations |
|
47
|
|
|
48
|
- **`.agents/skills/`** - User and agent-editable skills. Create new skills here. These can be freely modified, updated, and deleted. |
|
49
|
- **`~/.local/skills/`** - Replit-provided skills. These are read-only and managed by the platform. |
|
50
|
|
|
51
|
## Key Considerations |
|
52
|
|
|
53
|
- **Description is critical for discovery**: The description determines when you'll use this skill in the future. Be specific about triggers: |
|
54
|
- Bad: `description: Processes documents` (too vague) |
|
55
|
- Good: `description: Extracts text and images from PDF files. Use when the user asks to read, parse, or convert PDF documents.` |
|
56
|
- **Collaborate with the user**: Ask what they want included. Don't guess—the user knows what's important to them |
|
57
|
- **Include project-specific context**: File paths, naming conventions, API endpoints, preferred libraries—things unique to this user's workflow |
|
58
|
- **Use separate files for large content**: Organize reference material in folders (e.g., `.agents/skills/SKILLNAME/reference/api.md`) and link to it from SKILL.md |
|
59
|
- **Skills can be updated**: Skills in `.agents/skills/` are mutable. Update them as you discover better patterns—don't treat them as permanent. Iterate as the user's needs evolve |
|
60
|
- **Skills can reference other skills**: A skill can invoke or build upon other skills. Use this to compose complex workflows from simpler building blocks |
|
61
|
|
|
62
|
## Complete Example |
|
63
|
|
|
64
|
```markdown |
|
65
|
--- |
|
66
|
name: pr-review |
|
67
|
description: Reviews pull requests for code quality and security. Use when the user asks to review a PR or check code changes. |
|
68
|
--- |
|
69
|
|
|
70
|
# PR Review |
|
71
|
|
|
72
|
## Process |
|
73
|
|
|
74
|
1. Read the PR description and linked issues |
|
75
|
2. Review each file for issues |
|
76
|
3. Check for test coverage |
|
77
|
4. Look for security vulnerabilities |
|
78
|
5. Summarize findings with line references |
|
79
|
|
|
80
|
## What to Check |
|
81
|
|
|
82
|
- Logic errors and edge cases |
|
83
|
- Security issues (injection, XSS, auth bypass) |
|
84
|
- Performance concerns |
|
85
|
- Missing error handling |
|
86
|
|
|
87
|
## Output Format |
|
88
|
|
|
89
|
\`\`\`markdown |
|
90
|
## Summary |
|
91
|
[1-2 sentence overview] |
|
92
|
|
|
93
|
## Issues Found |
|
94
|
- **[severity]** file:line - description |
|
95
|
|
|
96
|
## Verdict |
|
97
|
[Approve / Request Changes / Comment] |
|
98
|
\`\`\` |
|
99
|
``` |