Write one skill.
Run it everywhere.
Authoring an AutoVault skill is writing a single SKILL.md that declares what it does, what it needs, and how it should be rendered for each calling agent. The vault handles the rest — validation, signing, scoping, transformation.
Anatomy of a SKILL.md
Hover a block to see why AutoVault cares about it. The body remains human-readable markdown; the frontmatter gives the vault enough structure to validate and render the skill safely.
1---2name: extract-pdf3version: 1.4.04description: "Extract structured text from PDF files."5author: autoworks-ai6license: Apache-2.078tools_required:9 - browser.fill_form10 - browser.click11 - fs.read12 - fs.write1314transformations:15 claude-code:16 browser.fill_form: chrome-devtools17 fs.read: read18 codex:19 browser.fill_form: browser_form20 fs.read: file_read2122permissions:23 network: false24 fs_scope: ["./inputs", "./outputs"]25 egress: deny2627---2829# Extract PDF text30Use this skill when the user provides a PDF path31and asks for its text contents, structure, or32summarization. Returns markdown with preserved33headings, lists, and table layout where possible.
Validation playground
The real gate runs deeper checks, but this playground shows the contract: structured metadata, canonical capabilities, permission declarations, and transformations get checked before anything is admitted.
Transformation manifest
Authors write against canonical capability names. AutoVault renders the skill for each caller at delivery time, so the author does not maintain platform forks.
Permissions
- Declare exact canonical tools.
- Declare network and filesystem scope.
- Keep secrets as named references only.
- Hide shell access inside prose.
- Embed credentials in frontmatter.
- Ship a generic helper skill with broad powers.