Skip to content

Commit 28205fe

Browse files
binary-huskybinary-husky
committed
update SKILL.md files with detailed descriptions for various AgentJet functionalities
1 parent c7ecec6 commit 28205fe

11 files changed

Lines changed: 304 additions & 7 deletions

File tree

‎ajet/copilot/auto-research-blueprint-execute-classic/SKILL.md‎

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,8 @@
1+
---
2+
name: auto-research-blueprint-execute-classic
3+
description: Execute AgentJet reinforcement learning experiments using experiment blueprints in classic (non-swarm) mode. Handles full lifecycle: launch experiment in tmux, monitor progress, analyze errors, collect results, and write finish flag. Use when the user wants to run AgentJet training experiments without the swarm distributed framework.
4+
---
5+
16
## 你的任务
27

38
1. 根据实验蓝图，运行实验

‎ajet/copilot/auto-research-blueprint-execute-swarm/SKILL.md‎

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,8 @@
1+
---
2+
name: auto-research-blueprint-execute-swarm
3+
description: Execute AgentJet reinforcement learning experiments using experiment blueprints in swarm mode. Handles full lifecycle: generate blueprint if needed, launch experiment in tmux, monitor progress, analyze errors, collect results, and write finish flag. Use when the user wants to run or debug AgentJet training experiments.
4+
---
5+
16
## 你的任务
27

38
0. 如果用户没有提供实验蓝图，则生成一个实验蓝图

‎ajet/copilot/conda-install-agentjet-swarm-server/SKILL.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
---
22
name: conda-install-agentjet-swarm-server
3-
description: Install agentjet swarm server with conda
3+
description: Install AgentJet swarm server using Conda. Handles Python 3.10 environment creation, dependency installation with the verl training backbone, flash-attn compilation, and optional PyPI mirror for China users.
44
license: Complete terms in LICENSE.txt
55
---
66

‎ajet/copilot/docker-install-agentjet-swarm-server/SKILL.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
---
22
name: docker-install-agentjet-swarm-server
3-
description: Install and run AgentJet Swarm Server via Docker with GPU support
3+
description: Install and run the AgentJet Swarm Server in a Docker container with NVIDIA GPU support. Use when the user wants to deploy a swarm server on a GPU machine via Docker, including GPU driver setup, Docker mirror configuration, model weight mounting, and server startup.
44
license: Complete terms in LICENSE.txt
55
---
66

‎ajet/copilot/download_from_swanlab_url/SKILL.md‎

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,8 @@
1+
---
2+
name: download-from-swanlab-url
3+
description: Download per-step time-series metric data (reward, entropy, response length, etc.) from a SwanLab cloud run URL as a pandas.DataFrame. Use when the user provides a SwanLab URL and wants to fetch or analyze training curves.
4+
---
5+
16
# Skill: Download metric data from a SwanLab run URL
27

38
## Goal

‎ajet/copilot/install-agentjet-client/SKILL.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
---
22
name: install-agentjet-client
3-
description: Install agentjet swarm server with uv package manager
3+
description: Install AgentJet client for connecting to a swarm server. Use when the user only needs to run the AgentJet client (not a swarm server) and does not need to run models locally, e.g. on a laptop. Installs basic requirements via `pip install -e .`.
44
license: Complete terms in LICENSE.txt
55
---
66

‎ajet/copilot/map-verl-config/SKILL.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
---
22
name: map-verl-config
3-
description: map verl config to agentjet config
3+
description: Map VERL training configuration to AgentJet configuration. Find VERL config in verl_default.yaml, check for existing mappings in config_auto_convertion_verl.jsonc, add new mappings to ajet_default.yaml and the conversion schema, and optionally add parameters to AgentJetJob.
44
license: Complete terms in LICENSE.txt
55
---
66

‎ajet/copilot/skill-normalizer/SKILL.md‎

Lines changed: 282 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,282 @@
1+
---
2+
name: skill-normalizer
3+
description: Convert skills in non-standard formats to the standard Agent Skills `SKILL.md` format. Validates YAML frontmatter (name, description, license, compatibility, metadata, allowed-tools), directory structure (SKILL.md, scripts/, references/, assets/), and best practices. Use when the user asks to normalize, validate, or fix a skill.
4+
license: Complete terms in LICENSE.txt
5+
---
6+
7+
Your task is to make skills in non-standard formats compatible with the Agent Skills ecosystem by converting them to the standard `SKILL.md` format. This document provides the specification for the `SKILL.md` format, including required and optional fields, directory structure, and best practices for writing effective skills.
8+
9+
---
10+
11+
> ## Documentation Index
12+
> Fetch the complete documentation index at: https://agentskills.io/llms.txt
13+
> Use this file to discover all available pages before exploring further.
14+
15+
# Specification
16+
17+
> The complete format specification for Agent Skills.
18+
19+
## Directory structure
20+
21+
A skill is a directory containing, at minimum, a `SKILL.md` file:
22+
23+
```
24+
skill-name/
25+
├── SKILL.md # Required: metadata + instructions
26+
├── scripts/ # Optional: executable code
27+
├── references/ # Optional: documentation
28+
├── assets/ # Optional: templates, resources
29+
└── ... # Any additional files or directories
30+
```
31+
32+
## `SKILL.md` format
33+
34+
The `SKILL.md` file must contain YAML frontmatter followed by Markdown content.
35+
36+
### Frontmatter
37+
38+
| Field | Required | Constraints |
39+
| --------------- | -------- | ----------------------------------------------------------------------------------------------------------------- |
40+
| `name` | Yes | Max 64 characters. Lowercase letters, numbers, and hyphens only. Must not start or end with a hyphen. |
41+
| `description` | Yes | Max 1024 characters. Non-empty. Describes what the skill does and when to use it. |
42+
| `license` | No | License name or reference to a bundled license file. |
43+
| `compatibility` | No | Max 500 characters. Indicates environment requirements (intended product, system packages, network access, etc.). |
44+
| `metadata` | No | Arbitrary key-value mapping for additional metadata. |
45+
| `allowed-tools` | No | Space-separated string of pre-approved tools the skill may use. (Experimental) |
46+
47+
<Card>
48+
**Minimal example:**
49+
50+
```markdown SKILL.md theme={null}
51+
---
52+
name: skill-name
53+
description: A description of what this skill does and when to use it.
54+
---
55+
```
56+
57+
**Example with optional fields:**
58+
59+
```markdown SKILL.md theme={null}
60+
---
61+
name: pdf-processing
62+
description: Extract PDF text, fill forms, merge files. Use when handling PDFs.
63+
license: Apache-2.0
64+
metadata:
65+
author: example-org
66+
version: "1.0"
67+
---
68+
```
69+
</Card>
70+
71+
#### `name` field
72+
73+
The required `name` field:
74+
75+
* Must be 1-64 characters
76+
* May only contain unicode lowercase alphanumeric characters (`a-z`) and hyphens (`-`)
77+
* Must not start or end with a hyphen (`-`)
78+
* Must not contain consecutive hyphens (`--`)
79+
* Must match the parent directory name
80+
81+
<Card>
82+
**Valid examples:**
83+
84+
```yaml theme={null}
85+
name: pdf-processing
86+
```
87+
88+
```yaml theme={null}
89+
name: data-analysis
90+
```
91+
92+
```yaml theme={null}
93+
name: code-review
94+
```
95+
96+
**Invalid examples:**
97+
98+
```yaml theme={null}
99+
name: PDF-Processing # uppercase not allowed
100+
```
101+
102+
```yaml theme={null}
103+
name: -pdf # cannot start with hyphen
104+
```
105+
106+
```yaml theme={null}
107+
name: pdf--processing # consecutive hyphens not allowed
108+
```
109+
</Card>
110+
111+
#### `description` field
112+
113+
The required `description` field:
114+
115+
* Must be 1-1024 characters
116+
* Should describe both what the skill does and when to use it
117+
* Should include specific keywords that help agents identify relevant tasks
118+
119+
<Card>
120+
**Good example:**
121+
122+
```yaml theme={null}
123+
description: Extracts text and tables from PDF files, fills PDF forms, and merges multiple PDFs. Use when working with PDF documents or when the user mentions PDFs, forms, or document extraction.
124+
```
125+
126+
**Poor example:**
127+
128+
```yaml theme={null}
129+
description: Helps with PDFs.
130+
```
131+
</Card>
132+
133+
#### `license` field
134+
135+
The optional `license` field:
136+
137+
* Specifies the license applied to the skill
138+
* We recommend keeping it short (either the name of a license or the name of a bundled license file)
139+
140+
<Card>
141+
**Example:**
142+
143+
```yaml theme={null}
144+
license: Proprietary. LICENSE.txt has complete terms
145+
```
146+
</Card>
147+
148+
#### `compatibility` field
149+
150+
The optional `compatibility` field:
151+
152+
* Must be 1-500 characters if provided
153+
* Should only be included if your skill has specific environment requirements
154+
* Can indicate intended product, required system packages, network access needs, etc.
155+
156+
<Card>
157+
**Examples:**
158+
159+
```yaml theme={null}
160+
compatibility: Designed for Claude Code (or similar products)
161+
```
162+
163+
```yaml theme={null}
164+
compatibility: Requires git, docker, jq, and access to the internet
165+
```
166+
167+
```yaml theme={null}
168+
compatibility: Requires Python 3.14+ and uv
169+
```
170+
</Card>
171+
172+
<Note>
173+
Most skills do not need the `compatibility` field.
174+
</Note>
175+
176+
#### `metadata` field
177+
178+
The optional `metadata` field:
179+
180+
* A map from string keys to string values
181+
* Clients can use this to store additional properties not defined by the Agent Skills spec
182+
* We recommend making your key names reasonably unique to avoid accidental conflicts
183+
184+
<Card>
185+
**Example:**
186+
187+
```yaml theme={null}
188+
metadata:
189+
author: example-org
190+
version: "1.0"
191+
```
192+
</Card>
193+
194+
#### `allowed-tools` field
195+
196+
The optional `allowed-tools` field:
197+
198+
* A space-separated string of tools that are pre-approved to run
199+
* Experimental. Support for this field may vary between agent implementations
200+
201+
<Card>
202+
**Example:**
203+
204+
```yaml theme={null}
205+
allowed-tools: Bash(git:*) Bash(jq:*) Read
206+
```
207+
</Card>
208+
209+
### Body content
210+
211+
The Markdown body after the frontmatter contains the skill instructions. There are no format restrictions. Write whatever helps agents perform the task effectively.
212+
213+
Recommended sections:
214+
215+
* Step-by-step instructions
216+
* Examples of inputs and outputs
217+
* Common edge cases
218+
219+
Note that the agent will load this entire file once it's decided to activate a skill. Consider splitting longer `SKILL.md` content into referenced files.
220+
221+
## Optional directories
222+
223+
### `scripts/`
224+
225+
Contains executable code that agents can run. Scripts should:
226+
227+
* Be self-contained or clearly document dependencies
228+
* Include helpful error messages
229+
* Handle edge cases gracefully
230+
231+
Supported languages depend on the agent implementation. Common options include Python, Bash, and JavaScript.
232+
233+
### `references/`
234+
235+
Contains additional documentation that agents can read when needed:
236+
237+
* `REFERENCE.md` - Detailed technical reference
238+
* `FORMS.md` - Form templates or structured data formats
239+
* Domain-specific files (`finance.md`, `legal.md`, etc.)
240+
241+
Keep individual [reference files](#file-references) focused. Agents load these on demand, so smaller files mean less use of context.
242+
243+
### `assets/`
244+
245+
Contains static resources:
246+
247+
* Templates (document templates, configuration templates)
248+
* Images (diagrams, examples)
249+
* Data files (lookup tables, schemas)
250+
251+
## Progressive disclosure
252+
253+
Agents load skills *progressively*, pulling in more detail only as a task calls for it. Skills should be structured to take advantage of this:
254+
255+
1. **Metadata** (\~100 tokens): The `name` and `description` fields are loaded at startup for all skills
256+
2. **Instructions** (\< 5000 tokens recommended): The full `SKILL.md` body is loaded when the skill is activated
257+
3. **Resources** (as needed): Files (e.g. those in `scripts/`, `references/`, or `assets/`) are loaded only when required
258+
259+
Keep your main `SKILL.md` under 500 lines. Move detailed reference material to separate files.
260+
261+
## File references
262+
263+
When referencing other files in your skill, use relative paths from the skill root:
264+
265+
```markdown SKILL.md theme={null}
266+
See [the reference guide](references/REFERENCE.md) for details.
267+
268+
Run the extraction script:
269+
scripts/extract.py
270+
```
271+
272+
Keep file references one level deep from `SKILL.md`. Avoid deeply nested reference chains.
273+
274+
## Validation
275+
276+
Use the [skills-ref](https://github.com/agentskills/agentskills/tree/main/skills-ref) reference library to validate your skills:
277+
278+
```bash theme={null}
279+
skills-ref validate ./my-skill
280+
```
281+
282+
This checks that your `SKILL.md` frontmatter is valid and follows all naming conventions.

‎ajet/copilot/train-complex-blackbox/SKILL.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
---
22
name: train-complex-blackbox
3-
description: Create a trainable agent loop or agent workflow with AgentJet
3+
description: Train complex blackbox agents (agents without clear reward signals) using AgentJet. Write dataset collectors, episode runners with LLM-as-Judge reward functions, and integrate with the AgentJet training loop.
44
license: Complete terms in LICENSE.txt
55
---
66

‎ajet/copilot/uv-install-agentjet-swarm-server/SKILL.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
---
22
name: uv-install-agentjet-swarm-server
3-
description: Install agentjet swarm server with uv package manager
3+
description: Install AgentJet swarm server using the UV package manager. Handles virtual environment creation with Python 3.10, dependency installation with the verl training backbone, flash-attn compilation, and optional PyPI mirror for China users.
44
license: Complete terms in LICENSE.txt
55
---
66

0 commit comments

Comments
 (0)