@jhustle44 / jcr-guide-pro
Code Commits Issues Pulls Cloud Agent CI/CD
copilot/create-zip-file-again
copilot/add-firebase copilot/change-app-name-to-jcrguru copilot/create-zip-file copilot/create-zip-file-again copilot/merge-any-updates main master
md 214 lines 6.57 KB
Rendered Raw
1 ---
2 name: skill-creator
3 description: Create new reusable skills to save instructions or workflows for future tasks.
4 ---
5
6 # Skill Creator
7
8 Create new skills that persist across sessions and teach future instances of yourself how to handle specific tasks, workflows, or conventions.
9
10 ## Process
11
12 ### 1. Capture Intent
13
14 Start by understanding what the user wants. If the current conversation already contains a workflow the user wants to capture (e.g., "turn this into a skill"), extract answers from the conversation history first — the tools used, the sequence of steps, corrections the user made, input/output formats observed. Then confirm before proceeding.
15
16 Key questions to answer (from conversation or by asking):
17
18 1. What should this skill enable you to do?
19 2. When should it trigger? (what phrases, contexts, or scenarios)
20
21 3. What is the expected output format?
22 4. Are there edge cases or special rules?
23
24 Adapt your communication to the user's technical level. Pay attention to context cues — explain terms if needed, but don't over-explain to technical users.
25
26 ### 2. Interview and Research
27
28 Proactively ask about edge cases, input/output formats, example files, success criteria, and dependencies. If the user has reference material or existing documentation, review it. Check if there are existing skills that this new one should reference or build upon.
29
30 ### 3. Choose a Name
31
32 Lowercase, hyphens only, max 64 characters (e.g., `deploy-checklist`,`api-conventions`).
33
34 ### 4. Write the SKILL.md
35
36 Place it at `.agents/skills/<skill-name>/SKILL.md`. For larger skills, organize supporting material in subdirectories.
37
38 ### 5. Confirm with the User
39
40 Summarize what the skill does, when it will activate, and walk through the key instructions. Iterate based on feedback.
41
42 ## Skill Directory Structure
43
44 ```text
45
46 skill-name/
47
48 ├── SKILL.md (required — main instructions)
49
50 ├── scripts/ (optional — executable code for deterministic tasks)
51
52 ├── references/ (optional — docs loaded into context as needed)
53
54 └── assets/ (optional — templates, icons, fonts used in output)
55
56 ```
57
58 ## Progressive Disclosure
59
60 Skills use a three-level loading system:
61
62 1. **Metadata** (name + description in frontmatter) — Always in context. This is how the skill gets discovered and triggered.
63 2. **SKILL.md body** — Loaded when the skill triggers. Keep under 500 lines.
64
65 3. **Bundled resources** (references/, scripts/, assets/) — Loaded as needed. For large reference files (>300 lines), include a table of contents. Reference them clearly from SKILL.md with guidance on when to read them.
66
67 ## SKILL.md Template
68
69 ```markdown
70
71 ---
72
73 name: <skill-name>
74
75 description: <What this skill does. When to use it. Max 1024 chars. Be specific about triggers.>
76
77 ---
78
79 # <Skill Title>
80
81 ## When to Use
82
83 - <Trigger condition 1>
84 - <Trigger condition 2>
85
86 ## Instructions
87
88 <Step-by-step instructions, conventions, or workflows.>
89
90 ## Examples
91
92 <Input/output pairs or concrete scenarios if applicable.>
93
94 ```
95
96 ## Writing Guide
97
98 ### Description is the Primary Trigger
99
100 The description determines when a future agent instance will load and follow this skill. It is the single most important line in the file. Be explicit about trigger phrases and scenarios. Lean toward being slightly "pushy" — skills tend to under-trigger rather than over-trigger.
101
102 - Bad: `description: Handles deployment`
103 - Good: `description: Runs the full deployment checklist before publishing. Use when the user asks to deploy, publish, go live, or ship to production.`
104
105 ### Writing Style
106
107 - Prefer the imperative form in instructions.
108 - Explain WHY things are important rather than relying on heavy-handed MUSTs. Help the future agent understand the reasoning.
109
110 - Use theory of mind — write instructions that guide toward good decisions in novel situations, not just narrow examples.
111 - Start by writing a draft, then review it with fresh eyes and improve it.
112
113 - Be concise — only include information a general-purpose agent would not already know. Skip general programming knowledge.
114
115 ### Specificity
116
117 Match specificity to fragility:
118
119 - **Exact commands and file paths** for things that break easily (build steps, config locations, migration commands).
120 - **General guidance** for flexible decisions (design choices, naming preferences).
121
122 ### Examples
123
124 Include examples when the skill involves structured output or non-obvious patterns. Format them clearly:
125
126 ```markdown
127
128 ## Commit message format
129
130 **Example 1:**
131
132 Input: Added user authentication with JWT tokens
133
134 Output: feat(auth): implement JWT-based authentication
135
136 ```
137
138 ### Output Format Definitions
139
140 When the skill produces structured output, define it explicitly:
141
142 ```markdown
143
144 ## Report structure
145
146 Always use this template:
147
148 # [Title]
149
150 ## Executive summary
151
152 ## Key findings
153
154 ## Recommendations
155
156 ```
157
158 ### Domain Organization
159
160 When a skill supports multiple domains or frameworks, organize by variant using reference files:
161
162 ```text
163
164 cloud-deploy/
165
166 ├── SKILL.md (workflow + selection logic)
167
168 └── references/
169
170 ├── aws.md
171
172 ├── gcp.md
173
174 └── azure.md
175
176 ```
177
178 The agent reads only the relevant reference file based on context.
179
180 ### Project-Specific Context
181
182 Include file paths, naming conventions, preferred libraries, API patterns — things unique to this workspace that a fresh agent instance would not know.
183
184 ### Composability
185
186 Skills can reference other skills. Use this to compose complex workflows from simpler building blocks when it makes sense.
187
188 ## Rules
189
190 - **Frontmatter is required.** Every SKILL.md must have `name`and`description` in YAML frontmatter.
191 - **Name constraints:** Max 64 chars, lowercase letters, numbers, and hyphens only.
192
193 - **Keep SKILL.md under 500 lines.** Use reference files for overflow.
194 - **Skills are mutable.** Update them as patterns evolve. They live in `.agents/skills/` and can be freely modified or deleted.
195
196 - **Skills must not be deceptive.** A skill's contents should not surprise the user in their intent if described.
197
198 ## Skill Location
199
200 All user-created skills go in `.agents/skills/<skill-name>/SKILL.md`. Platform-provided skills live in`.local/skills/` and are read-only.
201
202 ## Checklist Before Finishing
203
204 - [ ] Frontmatter has `name`and`description`
205 - [ ] Name is lowercase with hyphens only
206
207 - [ ] Description includes both WHAT and WHEN, and is specific enough to trigger reliably
208 - [ ] Instructions are clear enough for a fresh agent instance with no prior context
209
210 - [ ] Writing explains WHY, not just WHAT — reasoning over rules
211 - [ ] Large content is split into reference files, not crammed into SKILL.md
212
213 - [ ] File is saved to `.agents/skills/<skill-name>/SKILL.md`
214 - [ ] User has confirmed the skill content