Skip to content

Commit 817fd38

Browse files
sjnimsclaude
sjnims
and
claude
committed
feat(agent-development): add create-agent-skeleton.sh script
Add utility script to generate agent skeleton files with valid structure. The script validates agent names, creates properly formatted frontmatter and system prompt templates, and guides users through next steps. - Creates agent files with valid YAML frontmatter structure - Validates agent name format (lowercase, alphanumeric + hyphens) - Includes placeholder examples and system prompt sections - Provides helpful next-steps guidance after creation - References existing validate and test scripts Also updates SKILL.md to document the new script in Utility Scripts section and reference it in the Implementation Workflow. Fixes #26 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
1 parent b25446a commit 817fd38

2 files changed

Lines changed: 178 additions & 1 deletion

File tree

‎plugins/plugin-dev/skills/agent-development/SKILL.md‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -477,6 +477,7 @@ Working examples in `examples/`:
477477

478478
Development tools in `scripts/`:
479479

480+
- **`create-agent-skeleton.sh`** - Generate new agent file from template
480481
- **`validate-agent.sh`** - Validate agent file structure
481482
- **`test-agent-trigger.sh`** - Test if agent triggers correctly
482483

@@ -486,7 +487,7 @@ To create an agent for a plugin:
486487

487488
1. Define agent purpose and triggering conditions
488489
2. Choose creation method (AI-assisted or manual)
489-
3. Create `agents/agent-name.md` file
490+
3. Create agent file using skeleton: `./skills/agent-development/scripts/create-agent-skeleton.sh agent-name agents/`
490491
4. Write frontmatter with all required fields
491492
5. Write system prompt following best practices
492493
6. Include 2-4 triggering examples in description

‎plugins/plugin-dev/skills/agent-development/scripts/create-agent-skeleton.sh‎

Lines changed: 176 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,176 @@
1+
#!/bin/bash
2+
# Agent Skeleton Generator
3+
# Creates a new agent file with correct structure
4+
5+
set -euo pipefail
6+
7+
# Usage
8+
if [ $# -eq 0 ]; then
9+
echo "Usage: $0 <agent-name> [output-dir]"
10+
echo ""
11+
echo "Creates a skeleton agent file with:"
12+
echo " - Valid YAML frontmatter"
13+
echo " - Placeholder description with example block"
14+
echo " - Basic system prompt structure"
15+
echo ""
16+
echo "Arguments:"
17+
echo " agent-name Agent identifier (lowercase, numbers, hyphens)"
18+
echo " output-dir Directory to create file in (default: current directory)"
19+
echo ""
20+
echo "Examples:"
21+
echo " $0 code-reviewer agents/"
22+
echo " $0 test-generator"
23+
echo ""
24+
echo "After creation:"
25+
echo " 1. Edit the file to fill in placeholders"
26+
echo " 2. Add 2-4 triggering examples"
27+
echo " 3. Write detailed system prompt"
28+
echo " 4. Validate: ./scripts/validate-agent.sh <output-file>"
29+
echo " 5. Test triggers: ./scripts/test-agent-trigger.sh <output-file>"
30+
exit 1
31+
fi
32+
33+
AGENT_NAME="$1"
34+
OUTPUT_DIR="${2:-.}"
35+
OUTPUT_FILE="$OUTPUT_DIR/$AGENT_NAME.md"
36+
37+
echo "🔍 Creating agent skeleton: $AGENT_NAME"
38+
echo ""
39+
40+
# Validate name format (lowercase alphanumeric + hyphens, 3-50 chars)
41+
if ! [[ "$AGENT_NAME" =~ ^[a-z0-9][a-z0-9-]*[a-z0-9]$ ]]; then
42+
echo "❌ Invalid name: $AGENT_NAME"
43+
echo ""
44+
echo "Agent names must:"
45+
echo " - Start with a lowercase letter or number"
46+
echo " - End with a lowercase letter or number"
47+
echo " - Contain only lowercase letters, numbers, and hyphens"
48+
echo " - Be at least 3 characters long"
49+
echo ""
50+
echo "Valid examples: code-reviewer, test-gen, my-agent-123"
51+
echo "Invalid examples: Code-Reviewer, -agent, agent-, a"
52+
exit 1
53+
fi
54+
echo "✅ Name format valid"
55+
56+
# Validate name length
57+
name_length=${#AGENT_NAME}
58+
if [ "$name_length" -lt 3 ]; then
59+
echo "❌ Name too short: $name_length characters (minimum 3)"
60+
exit 1
61+
fi
62+
if [ "$name_length" -gt 50 ]; then
63+
echo "❌ Name too long: $name_length characters (maximum 50)"
64+
exit 1
65+
fi
66+
echo "✅ Name length valid ($name_length characters)"
67+
68+
# Check output directory exists
69+
if [ ! -d "$OUTPUT_DIR" ]; then
70+
echo "❌ Directory does not exist: $OUTPUT_DIR"
71+
echo ""
72+
echo "Create it first with: mkdir -p $OUTPUT_DIR"
73+
exit 1
74+
fi
75+
echo "✅ Output directory exists"
76+
77+
# Check file doesn't already exist
78+
if [ -f "$OUTPUT_FILE" ]; then
79+
echo "❌ File already exists: $OUTPUT_FILE"
80+
echo ""
81+
echo "Options:"
82+
echo " - Choose a different name"
83+
echo " - Delete the existing file: rm $OUTPUT_FILE"
84+
echo " - Edit the existing file directly"
85+
exit 1
86+
fi
87+
echo "✅ Output path available"
88+
89+
echo ""
90+
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
91+
echo "📝 Creating agent file..."
92+
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
93+
echo ""
94+
95+
# Create the agent skeleton file
96+
cat > "$OUTPUT_FILE" << EOF
97+
---
98+
name: $AGENT_NAME
99+
description: Use this agent when [describe triggering conditions]. Examples:
100+
101+
<example>
102+
Context: [Describe the situation]
103+
user: "[What the user says]"
104+
assistant: "[How Claude responds before triggering]"
105+
<commentary>
106+
[Why this agent should trigger]
107+
</commentary>
108+
assistant: "I'll use the $AGENT_NAME agent to [action]."
109+
</example>
110+
111+
<example>
112+
Context: [Second example situation]
113+
user: "[Another user request]"
114+
assistant: "[Claude's response]"
115+
<commentary>
116+
[Explanation of why this triggers the agent]
117+
</commentary>
118+
</example>
119+
120+
model: inherit
121+
color: blue
122+
---
123+
124+
You are [describe the agent's role and expertise].
125+
126+
**Your Core Responsibilities:**
127+
128+
1. [Primary responsibility]
129+
2. [Secondary responsibility]
130+
3. [Additional responsibility]
131+
132+
**Process:**
133+
134+
1. [First step]
135+
2. [Second step]
136+
3. [Third step]
137+
138+
**Quality Standards:**
139+
140+
- [Standard 1]
141+
- [Standard 2]
142+
- [Standard 3]
143+
144+
**Output Format:**
145+
146+
Provide results as:
147+
148+
- [What to include]
149+
- [How to structure]
150+
EOF
151+
152+
echo "✅ Created agent skeleton: $OUTPUT_FILE"
153+
echo ""
154+
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
155+
echo "📋 Next Steps"
156+
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
157+
echo ""
158+
echo "1. Edit the file to fill in placeholders:"
159+
echo " - Replace [bracketed text] with actual content"
160+
echo " - Update description to explain when agent triggers"
161+
echo " - Write 2-4 triggering examples"
162+
echo " - Develop detailed system prompt"
163+
echo ""
164+
echo "2. Choose appropriate settings:"
165+
echo " - model: inherit (default), sonnet, opus, or haiku"
166+
echo " - color: blue (default), cyan, green, yellow, magenta, or red"
167+
echo " - tools: Add array if restricting tools (optional)"
168+
echo ""
169+
echo "3. Validate the agent:"
170+
echo " ./scripts/validate-agent.sh $OUTPUT_FILE"
171+
echo ""
172+
echo "4. Test triggering:"
173+
echo " ./scripts/test-agent-trigger.sh $OUTPUT_FILE"
174+
echo ""
175+
echo "5. Load in Claude Code:"
176+
echo " cc --plugin-dir /path/to/plugin"

0 commit comments

Comments
 (0)