Suggestion
It would be useful to add a small Common tasks section to skills/linear-cli/SKILL.md (or as a dedicated references/common-tasks.md) with a handful of minimal, copy-pasteable recipes for the most frequent flows, e.g.:
- List issues for a project / team, filtered by state
- Create an issue (title + description from file)
- Update an issue's state / assignee / labels
- Add a comment from a file
- View an issue / get its URL
Why
Today the skill exposes a flat list of ~100 commands plus per-command reference files, which is great as a reference but doesn't surface the "how do I do the obvious thing" path. As an LLM agent reading the skill, it's easy to skip straight to linear api (GraphQL) for anything non-trivial — even though the skill itself recommends preferring the CLI — simply because there's no concrete CLI example for common filtered queries (e.g. "list active issues for a given project, sorted by updatedAt").
A short recipes section near the top would anchor the right tool choice before the agent commits to a heavier path.
Scope
Intentionally minimal: ~5–10 short snippets, one per task, no exhaustive flag coverage. The existing per-command reference files already handle the deep dive.
Suggestion
It would be useful to add a small Common tasks section to
skills/linear-cli/SKILL.md(or as a dedicatedreferences/common-tasks.md) with a handful of minimal, copy-pasteable recipes for the most frequent flows, e.g.:Why
Today the skill exposes a flat list of ~100 commands plus per-command reference files, which is great as a reference but doesn't surface the "how do I do the obvious thing" path. As an LLM agent reading the skill, it's easy to skip straight to
linear api(GraphQL) for anything non-trivial — even though the skill itself recommends preferring the CLI — simply because there's no concrete CLI example for common filtered queries (e.g. "list active issues for a given project, sorted by updatedAt").A short recipes section near the top would anchor the right tool choice before the agent commits to a heavier path.
Scope
Intentionally minimal: ~5–10 short snippets, one per task, no exhaustive flag coverage. The existing per-command reference files already handle the deep dive.