skill-forge-build

Name: skill-forge-build
Author: AgriciDaniel

78stars33forks

How to install skill-forge-build?

agricidaniel/skill-forge/skill-forge-build

$npx -y skills add AgriciDaniel/skill-forge --skill skill-forge-build

Installs into the current project.

Summary

Scaffold and build Claude Code skills from plans or descriptions. Generates SKILL.md files, sub-skills, scripts, references, agents, and templates following the Agent Skills standard. Use when user says "build skill", "scaffold skill", "generate skill", "create SKILL.md", or "imp

Files · 1

View on GitHub

SKILL.md

1	# Skill Builder — Scaffold & Generate
2
3	## Process
4
5	### Step 1: Gather Inputs
6
7	Determine what we're building from:
8	- Plan document: Output from `/skill-forge plan` (preferred)
9	- Description: Natural language description of the skill
10	- Existing skill: Path to a skill to use as foundation
11
12	If no plan exists, run a quick planning pass (ask 3 key questions):
13	1. What does the skill do? (1-2 sentences)
14	2. What commands should it have? (list)
15	3. Does it need scripts or external tools? (yes/no)
16
17	### Step 2: Generate Frontmatter
18
19	The most critical step. The description field determines activation.
20
21	Framework for writing descriptions:
22
23	```
24	[Capability statement] + [Detailed capabilities] + [Trigger phrases]
25	```
26
27	Rules:
28	- Under 1024 characters total
29	- No XML angle brackets (< >)
30	- Include 5-10 trigger phrases users would say
31	- Mention file types if relevant
32	- Add negative triggers if risk of over-triggering
33
34	Use `references/description-guide.md` for the full framework.
35
36	### Step 3: Write Main SKILL.md
37
38	Structure the body following this template:
39
40	```markdown
41	# [Skill Name] -- [Tagline]
42
43	[1-2 sentence overview]
44
45	## Quick Reference
46
47	\| Command \| What it does \|
48	\|---------\|-------------\|
49	\| /name cmd1 \| Description \|
50	\| /name cmd2 \| Description \|
51
52	## Orchestration Logic
53
54	[How commands route to sub-skills]
55	[Decision trees for conditional routing]
56
57	## [Domain-Specific Section]
58
59	[Core rules, thresholds, quality gates]
60
61	## Reference Files
62
63	[List of on-demand reference files]
64
65	## Sub-Skills
66
67	[List of sub-skills with one-line descriptions]
68	```
69
70	Critical SKILL.md rules:
71	- Under 500 lines
72	- Under 5000 tokens
73	- Actionable instructions (not vague)
74	- Include error handling
75	- Link references, don't inline them
76
77	### Step 4: Generate Sub-Skills (Tier 3-4)
78
79	For each sub-skill, generate a SKILL.md with:
80
81	```yaml
82	---
83	name: parent-child
84	description: >
85	[What it does]. Use when user says "[trigger 1]", "[trigger 2]",
86	"[trigger 3]", or "[trigger 4]".
87	---
88	```
89
90	Sub-skill body guidelines:
91	- Self-contained instructions for ONE workflow
92	- Cross-reference other sub-skills by name (not inline their content)
93	- Reference shared files in the parent skill's `references/` directory
94	- Include explicit input/output definitions
95
96	### Step 5: Generate Scripts (Tier 2-4)
97
98	For each script:
99
100	```python
101	#!/usr/bin/env python3
102	"""
103	Purpose: [What this script does]
104	Input: [Expected input format]
105	Output: [Expected output format]
106	Usage: python scripts/script_name.py [args]
107	"""
108
109	import argparse
110	import json
111	import sys
112	from typing import Any
113
114
115	def main(args: argparse.Namespace) -> dict[str, Any]:
116	"""Main execution function."""
117	# Validate inputs
118	# Do the work
119	# Return structured output
120	pass
121
122
123	if __name__ == "__main__":
124	parser = argparse.ArgumentParser(description="[Script description]")
125	parser.add_argument("input", help="[Input description]")
126	parser.add_argument("--output", "-o", help="Output file path")
127	args = parser.parse_args()
128
129	try:
130	result = main(args)
131	print(json.dumps(result, indent=2))
132	except Exception as e:
133	print(json.dumps({"error": str(e)}), file=sys.stderr)
134	sys.exit(1)
135	```
136
137	Script rules:
138	- One script, one responsibility
139	- CLI interface with argparse
140	- Structured JSON output
141	- Clear error messages
142	- No dependencies beyond stdlib when possible
143
144	### Step 6: Generate Reference Files
145
146	For each reference file:
147	- Focus on ONE topic
148	- Under 200 lines
149	- Include concrete examples
150	- Use tables for thresholds and specifications
151	- Mark with "Load on-demand" in parent SKILL.md
152
153	### Step 7: Generate Agent Definitions (Tier 4)
154
155	Agent files use YAML frontmatter (different from skills):
156
157	```markdown
158	---
159	name: [skill-name]-[role]
160	description: >
161	[What this agent analyzes].
162	<example>User says: "[trigger]"</example>
163	model: inherit
164	color: blue
165	tools:
166	- Read
167	- Grep
168	- Glob
169	---
170
171	You are a [role] specialist.
172
173	## Your Role
174	[What this agent does]
175
176	## Process
177	1. [Steps...]
178
179	## Output Format
180	Return structured markdown with findings and recommendations.
181	```
182
183	Key differences from skills: Agent `name` is 3-50 chars, descriptions CAN use
184	`<example>` blocks, agent-only fields include `tools`, `color`, `permissionMode`,
185	`maxTurns`, `memory`. Body becomes the system prompt (second person).
186
187	### Step 8: Generate Install Script
188
189	Create `install.sh` that copies files to `~/.claude/skills/` and `~/.claude/agents/`.
190
191	### Step 9: Validate
192
193	Run `python scripts/validate_skill.py <path>` on the generated skill.
194	Check all quality gates from the main SKILL.md.
195
196	### Step 10: Output Summary
197
198	Present the user with:
199	1. Generated file tree
200	2. Key f

People who liked this also viewed

6 picks

Type

skill-forge-build

78stars33forks

View on GitHub

How to install skill-forge-build?

agricidaniel/skill-forge/skill-forge-build

$npx -y skills add AgriciDaniel/skill-forge --skill skill-forge-build

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Skill Builder — Scaffold & Generate
2
3	## Process
4
5	### Step 1: Gather Inputs
6
7	Determine what we're building from:
8	- Plan document: Output from `/skill-forge plan` (preferred)
9	- Description: Natural language description of the skill
10	- Existing skill: Path to a skill to use as foundation
11
12	If no plan exists, run a quick planning pass (ask 3 key questions):
13	1. What does the skill do? (1-2 sentences)
14	2. What commands should it have? (list)
15	3. Does it need scripts or external tools? (yes/no)
16
17	### Step 2: Generate Frontmatter
18
19	The most critical step. The description field determines activation.
20
21	Framework for writing descriptions:
22
23	```
24	[Capability statement] + [Detailed capabilities] + [Trigger phrases]
25	```
26
27	Rules:
28	- Under 1024 characters total
29	- No XML angle brackets (< >)
30	- Include 5-10 trigger phrases users would say
31	- Mention file types if relevant
32	- Add negative triggers if risk of over-triggering
33
34	Use `references/description-guide.md` for the full framework.
35
36	### Step 3: Write Main SKILL.md
37
38	Structure the body following this template:
39
40	```markdown
41	# [Skill Name] -- [Tagline]
42
43	[1-2 sentence overview]
44
45	## Quick Reference
46
47	\| Command \| What it does \|
48	\|---------\|-------------\|
49	\| /name cmd1 \| Description \|
50	\| /name cmd2 \| Description \|
51
52	## Orchestration Logic
53
54	[How commands route to sub-skills]
55	[Decision trees for conditional routing]
56
57	## [Domain-Specific Section]
58
59	[Core rules, thresholds, quality gates]
60
61	## Reference Files
62
63	[List of on-demand reference files]
64
65	## Sub-Skills
66
67	[List of sub-skills with one-line descriptions]
68	```
69
70	Critical SKILL.md rules:
71	- Under 500 lines
72	- Under 5000 tokens
73	- Actionable instructions (not vague)
74	- Include error handling
75	- Link references, don't inline them
76
77	### Step 4: Generate Sub-Skills (Tier 3-4)
78
79	For each sub-skill, generate a SKILL.md with:
80
81	```yaml
82	---
83	name: parent-child
84	description: >
85	[What it does]. Use when user says "[trigger 1]", "[trigger 2]",
86	"[trigger 3]", or "[trigger 4]".
87	---
88	```
89
90	Sub-skill body guidelines:
91	- Self-contained instructions for ONE workflow
92	- Cross-reference other sub-skills by name (not inline their content)
93	- Reference shared files in the parent skill's `references/` directory
94	- Include explicit input/output definitions
95
96	### Step 5: Generate Scripts (Tier 2-4)
97
98	For each script:
99
100	```python
101	#!/usr/bin/env python3
102	"""
103	Purpose: [What this script does]
104	Input: [Expected input format]
105	Output: [Expected output format]
106	Usage: python scripts/script_name.py [args]
107	"""
108
109	import argparse
110	import json
111	import sys
112	from typing import Any
113
114
115	def main(args: argparse.Namespace) -> dict[str, Any]:
116	"""Main execution function."""
117	# Validate inputs
118	# Do the work
119	# Return structured output
120	pass
121
122
123	if __name__ == "__main__":
124	parser = argparse.ArgumentParser(description="[Script description]")
125	parser.add_argument("input", help="[Input description]")
126	parser.add_argument("--output", "-o", help="Output file path")
127	args = parser.parse_args()
128
129	try:
130	result = main(args)
131	print(json.dumps(result, indent=2))
132	except Exception as e:
133	print(json.dumps({"error": str(e)}), file=sys.stderr)
134	sys.exit(1)
135	```
136
137	Script rules:
138	- One script, one responsibility
139	- CLI interface with argparse
140	- Structured JSON output
141	- Clear error messages
142	- No dependencies beyond stdlib when possible
143
144	### Step 6: Generate Reference Files
145
146	For each reference file:
147	- Focus on ONE topic
148	- Under 200 lines
149	- Include concrete examples
150	- Use tables for thresholds and specifications
151	- Mark with "Load on-demand" in parent SKILL.md
152
153	### Step 7: Generate Agent Definitions (Tier 4)
154
155	Agent files use YAML frontmatter (different from skills):
156
157	```markdown
158	---
159	name: [skill-name]-[role]
160	description: >
161	[What this agent analyzes].
162	<example>User says: "[trigger]"</example>
163	model: inherit
164	color: blue
165	tools:
166	- Read
167	- Grep
168	- Glob
169	---
170
171	You are a [role] specialist.
172
173	## Your Role
174	[What this agent does]
175
176	## Process
177	1. [Steps...]
178
179	## Output Format
180	Return structured markdown with findings and recommendations.
181	```
182
183	Key differences from skills: Agent `name` is 3-50 chars, descriptions CAN use
184	`<example>` blocks, agent-only fields include `tools`, `color`, `permissionMode`,
185	`maxTurns`, `memory`. Body becomes the system prompt (second person).
186
187	### Step 8: Generate Install Script
188
189	Create `install.sh` that copies files to `~/.claude/skills/` and `~/.claude/agents/`.
190
191	### Step 9: Validate
192
193	Run `python scripts/validate_skill.py <path>` on the generated skill.
194	Check all quality gates from the main SKILL.md.
195
196	### Step 10: Output Summary
197
198	Present the user with:
199	1. Generated file tree
200	2. Key f

People who liked this also viewed

6 picks

Type