Question
The $ picker is alphabetical, so with 100+ skills Api Design and Archive Loop Scaffolding
outrank whatever you came for. ask-matt solves this already — but as structure, not
intelligence: a main flow (idea → ship), on-ramps (triage, diagnosing-bugs, wayfinder), codebase
health, and a vocabulary layer underneath.
Decided: render that structure in the picker, sourced from an optional manifest file in the skills
directory — not hardcoded in T3, not squeezed into per-skill frontmatter (frontmatter cannot
express a flow; ordering and branching live between skills).
Open: what does the manifest look like?
- Format and filename. Where exactly does it sit relative to the skills it groups?
- Can it express the shapes
ask-matt needs — ordered main flow, on-ramps that merge onto it,
standalone, and an underneath layer?
- What happens when a skill is in the directory but not the manifest, or in the manifest but missing?
- How do user-scope and project-scope manifests compose, given skills already resolve
most-specific-wins?
- What does the picker show when there is no manifest at all? (Must degrade to today's behaviour.)
- Does the manifest belong to Matt's repo, or is it a T3 convention that his repo happens to adopt?
Question
The
$picker is alphabetical, so with 100+ skillsApi DesignandArchive Loop Scaffoldingoutrank whatever you came for.
ask-mattsolves this already — but as structure, notintelligence: a main flow (idea → ship), on-ramps (triage, diagnosing-bugs, wayfinder), codebase
health, and a vocabulary layer underneath.
Decided: render that structure in the picker, sourced from an optional manifest file in the skills
directory — not hardcoded in T3, not squeezed into per-skill frontmatter (frontmatter cannot
express a flow; ordering and branching live between skills).
Open: what does the manifest look like?
ask-mattneeds — ordered main flow, on-ramps that merge onto it,standalone, and an underneath layer?
most-specific-wins?