120 lines
4.0 KiB
Markdown
120 lines
4.0 KiB
Markdown
# skvlt
|
|
|
|
**_[汉语](./README.zh.md)_**
|
|
|
|
This repository is organized around a single-file Agent Skills manifest, [`skvlt.yaml`](./skvlt.yaml), for [Skills Vault](https://github.com/xixu-me/skills-vault).
|
|
|
|
It is intended for people who want a reviewed, portable snapshot of approved skill sources and installed skill names.
|
|
|
|
## Repository Contents
|
|
|
|
- [`skvlt.yaml`](./skvlt.yaml): the source-of-truth manifest
|
|
- [`MANIFEST_POLICY.md`](./MANIFEST_POLICY.md): the agent-facing policy for additions, replacements, rejections, and escalations
|
|
|
|
> [!IMPORTANT]
|
|
> In normal use, manifest changes are proposed and applied by agents following the policy.
|
|
|
|
## Manifest Overview
|
|
|
|
The manifest records:
|
|
|
|
- approved upstream skill sources
|
|
- installed skill names under each source
|
|
- integrity fields such as `total_sources`, `total_skills`, and per-source `count`
|
|
- the current manifest scope, `global`
|
|
|
|
Minimal example:
|
|
|
|
```yaml
|
|
total_sources: 21
|
|
total_skills: 137
|
|
scope: "global"
|
|
|
|
sources:
|
|
"openai/skills":
|
|
count: 27
|
|
skills:
|
|
- "openai-docs"
|
|
- "slides"
|
|
- "sentry"
|
|
```
|
|
|
|
Today, the manifest covers a broad set of real-world development tasks, including general engineering workflows, AI, cloud platforms, security, testing, documentation, design, deployment, and research-oriented work.
|
|
|
|
It is especially useful for:
|
|
|
|
- developers working across a wide range of modern software tasks
|
|
- teams that want a shared, reviewed skills baseline across domains
|
|
- Skills Vault users who do not want to assemble trusted sources one by one
|
|
|
|
## Using It With Skills Vault
|
|
|
|
This repository is a maintained manifest source intended to be used with Skills Vault.
|
|
|
|
Recommended workflow:
|
|
|
|
1. [Fork this repository](https://github.com/xixu-me/skvlt/fork) and clone it.
|
|
2. Run the restore flow through Skills Vault.
|
|
|
|
Example:
|
|
|
|
```bash
|
|
git clone https://github.com/xixu-me/skvlt.git
|
|
cd skvlt
|
|
bunx skvlt restore --all
|
|
```
|
|
|
|
This repository is not a standalone installer. Its role is to provide a maintained manifest for the Skills Vault toolchain.
|
|
|
|
> [!IMPORTANT]
|
|
> This repository is meant to be used with Skills Vault. Standalone use outside the Skills Vault workflow is not considered a supported primary workflow.
|
|
|
|
## Formatting
|
|
|
|
This repository now includes a lightweight Prettier-based formatting toolchain for Markdown and YAML files.
|
|
|
|
Local commands:
|
|
|
|
```bash
|
|
bun install
|
|
bun run format:check
|
|
bun run format
|
|
```
|
|
|
|
GitHub Actions workflows:
|
|
|
|
- `Format Check`: runs formatting validation on pull requests and pushes to `main`
|
|
- `Format Fix`: manual `workflow_dispatch` workflow that formats the selected branch and pushes a `chore: apply formatting` commit only when changes are needed
|
|
|
|
## Change Policy
|
|
|
|
Because this manifest is global rather than project-specific, changes are handled conservatively.
|
|
|
|
High-level rules include:
|
|
|
|
- prefer small, auditable changes
|
|
- prefer adding skills under existing trusted sources over introducing source churn
|
|
- reject skills tied to a specific repository, agent, or runtime unless explicitly allowed
|
|
- escalate ambiguous overlap, process-shaping changes, and source-level edits to a human reviewer
|
|
|
|
See [`MANIFEST_POLICY.md`](./MANIFEST_POLICY.md) for the complete decision model.
|
|
|
|
## Updating The Manifest Through Agents
|
|
|
|
In most cases, manifest changes should be requested through an agent rather than edited directly in [`skvlt.yaml`](./skvlt.yaml).
|
|
|
|
Recommended flow:
|
|
|
|
1. [Fork this repository](https://github.com/xixu-me/skvlt/fork) and clone it.
|
|
2. Give an agent access to the cloned repository and request a manifest change.
|
|
3. Provide the candidate skill or source and the change you want.
|
|
4. Let the agent evaluate the request against [`MANIFEST_POLICY.md`](./MANIFEST_POLICY.md) and prepare a patch if the change is justified.
|
|
5. Review the agent's rationale and diff before accepting the change.
|
|
6. Commit the result.
|
|
|
|
If the case is ambiguous, high-impact, or source-level, the agent should normally escalate to a human instead of applying the change automatically.
|
|
|
|
## License
|
|
|
|
Released under the MIT License. See [`LICENSE`](./LICENSE) for details.
|