author-flowglad-skill
Installation
SKILL.md
Author Flowglad Skill
Create a durable Flowglad Skill package and follow its immutable revision lifecycle.
Authoring process
- Gather the reusable behavior, inputs, outputs, failure conditions, and representative synthetic cases.
- Use
skills.listandskills.getwhen Flowglad MCP is available. Prefer a suitable active Skill over creating a duplicate. - For a new Skill, derive the slug from the request. For an update, resolve the exact Skill and base revision; never infer identities from names.
- Read
references/PACKAGE_CONTRACT.mdcompletely. - Copy
assets/starter-package, renameSKILL.template.mdto exactlySKILL.md, replace every placeholder, and extend only what the capability needs. - For every selected Python function, read
references/OWS_COMPATIBILITY.mdand use the OWS code ABI. Do not add a CLI wrapper around OWS code. - Create or revise
SKILL.md, support files,scripts/**.py, and smallflowglad-ows-code-test-v1cases undertests/**.json. Keep scratch work outside the package. - Run Python syntax checks and any available Flowglad OWS verifier. Do not execute the selected function as a positional-argument CLI.
- Run this skill's bundled helper as
python3 /path/to/author-flowglad-skill/scripts/build-package-payload.py PACKAGE_DIRfor inline MCP preflight. For large packages or MCP byte-transport failures, add--transport stagedand followreferences/MCP_LIFECYCLE.md. - Submit only when the user explicitly asks to create, upload, publish, or update the remote Skill.
- Poll
skills.get_lifecycle_request, thenskills.getuntil the candidate reaches its final verification state. Repair rejected authored bytes as a new immutable candidate.