Write project playbooks with AI
Document the way a repository already creates a component or module, with evidence for each command and room for unresolved conventions.
You can write project playbooks with AI when the repository already knows how to do a job that newcomers keep rebuilding by hand. Give playbook-authoring one recurring procedure, such as adding a package with the team's own generator. The first useful output is a list of candidates and the local files that support them. No document needs to be written at that point.
A task name declared in a package manifest supplies stronger evidence than a folder that happens to resemble the last component. The guide distinguishes configured steps from observed patterns and asks for unresolved entries to be checked. It also leaves out a single command dressed up as a new procedure. The result is easier to maintain when a generator is renamed or a workspace splits.
What it does
- Locate a recurring task across a repository's package manifests.
- Trace a registered generator to its configuration before documenting it.
- Distinguish an actual build task from a naming convention seen in existing files.
- Discard proposed guides that would simply restate one obvious command.
- Review which instructions become stale after a task declaration changes.
Install and use
- Use a file-capable agent with access to an authorised repository. Consumer projects follow the manual procedure; the maintainer's derivation script is separate.
- Inspect the candidate list and check the file supporting each asserted command.
- Write only the accepted procedures, then reopen each guide and verify its references.
FAQ
Source
Made by event4u. Original skill: playbook-authoring, published on GitHub under the MIT licence.
View on GitHub