skill-creator

Name: skill-creator
Author: pssah4

215stars23forks

How to install skill-creator?

pssah4/vault-operator/skill-creator

$npx -y skills add pssah4/vault-operator --skill skill-creator

Installs into the current project.

Summary

Guide for creating effective Vault Operator skills. Use when the user wants to build a new skill (or refine an existing one) that captures a repeatable workflow, a domain expertise, or a tool-integration recipe. Triggers include phrases like "build me a skill", "kannst du das aut

Files · 1

View on GitHub

SKILL.md

1	# Skill Creator
2
3	This skill guides the agent through creating a Vault Operator skill that another agent run can pick up and execute reliably. It is the basic authoring flow: initialize, edit, validate.
4
5	If the `skill-creator-pro` skill is available, prefer it for anything beyond a trivial one-step skill. The Pro flow adds a guided requirements dialog (it interviews the user when they cannot state exactly what they need), a feasibility check against the actual plugin runtime and sandbox limits, a confirmed skill brief before any file is written, and a dry-run. This basic flow stays fully usable on its own; the Pro flow is the upgrade for designing non-trivial skills reliably.
6
7	## Binding rules (read first, no exceptions)
8
9	1. The skill folder lives ONLY at `<agent-folder>/data/skills/{skill_name}/`. Any other path (`Tools & Settings/Skills/`, `Notes/skills/`, vault-root, etc.) is invisible to the discovery layer. A skill written elsewhere does not exist.
10	2. `init_skill` is the ONLY way to create the folder. Do NOT call `create_folder` or `write_file` to seed a new skill manually. Even when you "know" the layout, route through `init_skill` so the validator stays happy and the source-frontmatter is correct.
11	3. Existing artifacts at the wrong path are NOT proof the task is done. If you find a half-finished `weekly-meeting-summary` under `Tools & Settings/Skills/` or similar, treat it as garbage from a previous broken run. Delete it after you have the real skill landed at `data/skills/{name}/`.
12	4. `quick_validate` is mandatory before declaring done. Step 5 is not optional. A skill that "looks right" but was never validated is not finished.
13	5. Do not short-circuit by reusing a similar-sounding existing skill. When the user asks to BUILD a new skill, build a new one. Reuse means edit-and-save under the same name -- different name means new skill, even if the domain overlaps with something that already exists.
14
15	## About Vault Operator skills
16
17	A Vault Operator skill is a folder under `<agent-folder>/data/skills/{skill_name}/` that contains:
18
19	- `SKILL.md` -- the only required file. YAML frontmatter (`name`, `description`) plus markdown body.
20	- `scripts/` -- optional JavaScript helpers executable via `run_skill_script`.
21	- `references/` -- optional markdown files the agent loads on demand.
22	- `assets/` -- optional binary files the skill emits as output.
23
24	The frontmatter `description` is the trigger. The agent picks the skill based on description alone, never reads the body until the skill activates.
25
26	## Core principles
27
28	### Concise wins
29
30	Context window is a shared resource. Only add information the agent does not already have. Every paragraph must justify its tokens. Prefer one concrete example over a paragraph of theory.
31
32	### Match freedom to fragility
33
34	Pick the right level of detail for each step:
35
36	- High freedom (text instructions) -- many valid approaches, agent decides.
37	- Medium freedom (pseudocode, parametrized scripts) -- preferred pattern with variation room.
38	- Low freedom (specific scripts, few parameters) -- fragile or sequence-critical work.
39
40	A narrow bridge with cliffs needs guardrails. An open field allows many routes.
41
42	### Progressive disclosure
43
44	Keep `SKILL.md` under 500 lines. When a section grows past that, move it into a `references/` file and link to it from the body. Reference files load on demand; the body loads on every skill activation.
45
46	## Anatomy of a skill folder
47
48	```
49	<agent-folder>/data/skills/{skill_name}/
50	├── SKILL.md (required)
51	├── scripts/ (optional)
52	│ └── *.js (executable via run_skill_script)
53	├── references/ (optional, load on demand via read_file)
54	│ └── *.md
55	└── assets/ (optional, copied into output)
56	└── *
57	```
58
59	### Frontmatter rules (binding)
60
61	Only two fields are written:
62
63	```yaml
64	---
65	name: my-skill
66	description: One concise paragraph that explains WHAT the skill does and WHEN to use it. Include explicit triggers so the agent picks it up from the user phrasing.
67	---
68	```
69
70	Hard constraints enforced by the validator:
71
72	- `name` is kebab-case (`[a-z0-9-]+`), max 64 characters, no leading or trailing hyphen, no double hyphens.
73	- `name` cannot contain the reserved words `anthropic` or `claude`.
74	- `description` is max 1024 characters, no angle brackets, single line.
75	- Allowed-only properties: `name`, `description`. Adding extra keys triggers a warning in the discovery layer.
76
77	### Anti-patterns
78
79	Do NOT create the following in a skill folder:
80
81	- `README.md` -- the skill body already tells the agent everything it needs.

People who liked this also viewed

6 picks

Type

skill-creator

215stars23forks

View on GitHub

How to install skill-creator?

pssah4/vault-operator/skill-creator

$npx -y skills add pssah4/vault-operator --skill skill-creator

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Skill Creator
2
3	This skill guides the agent through creating a Vault Operator skill that another agent run can pick up and execute reliably. It is the basic authoring flow: initialize, edit, validate.
4
5	If the `skill-creator-pro` skill is available, prefer it for anything beyond a trivial one-step skill. The Pro flow adds a guided requirements dialog (it interviews the user when they cannot state exactly what they need), a feasibility check against the actual plugin runtime and sandbox limits, a confirmed skill brief before any file is written, and a dry-run. This basic flow stays fully usable on its own; the Pro flow is the upgrade for designing non-trivial skills reliably.
6
7	## Binding rules (read first, no exceptions)
8
9	1. The skill folder lives ONLY at `<agent-folder>/data/skills/{skill_name}/`. Any other path (`Tools & Settings/Skills/`, `Notes/skills/`, vault-root, etc.) is invisible to the discovery layer. A skill written elsewhere does not exist.
10	2. `init_skill` is the ONLY way to create the folder. Do NOT call `create_folder` or `write_file` to seed a new skill manually. Even when you "know" the layout, route through `init_skill` so the validator stays happy and the source-frontmatter is correct.
11	3. Existing artifacts at the wrong path are NOT proof the task is done. If you find a half-finished `weekly-meeting-summary` under `Tools & Settings/Skills/` or similar, treat it as garbage from a previous broken run. Delete it after you have the real skill landed at `data/skills/{name}/`.
12	4. `quick_validate` is mandatory before declaring done. Step 5 is not optional. A skill that "looks right" but was never validated is not finished.
13	5. Do not short-circuit by reusing a similar-sounding existing skill. When the user asks to BUILD a new skill, build a new one. Reuse means edit-and-save under the same name -- different name means new skill, even if the domain overlaps with something that already exists.
14
15	## About Vault Operator skills
16
17	A Vault Operator skill is a folder under `<agent-folder>/data/skills/{skill_name}/` that contains:
18
19	- `SKILL.md` -- the only required file. YAML frontmatter (`name`, `description`) plus markdown body.
20	- `scripts/` -- optional JavaScript helpers executable via `run_skill_script`.
21	- `references/` -- optional markdown files the agent loads on demand.
22	- `assets/` -- optional binary files the skill emits as output.
23
24	The frontmatter `description` is the trigger. The agent picks the skill based on description alone, never reads the body until the skill activates.
25
26	## Core principles
27
28	### Concise wins
29
30	Context window is a shared resource. Only add information the agent does not already have. Every paragraph must justify its tokens. Prefer one concrete example over a paragraph of theory.
31
32	### Match freedom to fragility
33
34	Pick the right level of detail for each step:
35
36	- High freedom (text instructions) -- many valid approaches, agent decides.
37	- Medium freedom (pseudocode, parametrized scripts) -- preferred pattern with variation room.
38	- Low freedom (specific scripts, few parameters) -- fragile or sequence-critical work.
39
40	A narrow bridge with cliffs needs guardrails. An open field allows many routes.
41
42	### Progressive disclosure
43
44	Keep `SKILL.md` under 500 lines. When a section grows past that, move it into a `references/` file and link to it from the body. Reference files load on demand; the body loads on every skill activation.
45
46	## Anatomy of a skill folder
47
48	```
49	<agent-folder>/data/skills/{skill_name}/
50	├── SKILL.md (required)
51	├── scripts/ (optional)
52	│ └── *.js (executable via run_skill_script)
53	├── references/ (optional, load on demand via read_file)
54	│ └── *.md
55	└── assets/ (optional, copied into output)
56	└── *
57	```
58
59	### Frontmatter rules (binding)
60
61	Only two fields are written:
62
63	```yaml
64	---
65	name: my-skill
66	description: One concise paragraph that explains WHAT the skill does and WHEN to use it. Include explicit triggers so the agent picks it up from the user phrasing.
67	---
68	```
69
70	Hard constraints enforced by the validator:
71
72	- `name` is kebab-case (`[a-z0-9-]+`), max 64 characters, no leading or trailing hyphen, no double hyphens.
73	- `name` cannot contain the reserved words `anthropic` or `claude`.
74	- `description` is max 1024 characters, no angle brackets, single line.
75	- Allowed-only properties: `name`, `description`. Adding extra keys triggers a warning in the discovery layer.
76
77	### Anti-patterns
78
79	Do NOT create the following in a skill folder:
80
81	- `README.md` -- the skill body already tells the agent everything it needs.

People who liked this also viewed

6 picks

Type