weside-ai

standup

Where this branch stands: ticket, PR, CI, what changed, what comes next and whether you must act — read-only, one screen. Triggers: "/we:standup", "where am I", "what's the state".

weside-ai 0 Updated 3w ago
GitHub

Install

npx skillscat add weside-ai/claude-code-plugin/standup

Install via the SkillsCat registry.

SKILL.md

/we:standup

Read-only: read git, the plan, the ticket and gh; write nothing, dispatch nobody, move no ticket.
The reader comes back to this window after an hour elsewhere and needs three answers: what is now
true that was not, which story comes next (by key), and what the human must do (often nothing).
When the user's instruction files define a status form
(for example a sentence cap), that form wins over the layout below.

Gather

  1. git status -sb: branch, ahead/behind, dirty files.
  2. Key from the branch (<type>/<KEY>-…) → docs/plans/<KEY>-story.md: frontmatter status and
    its ### Phase N blocks; the ticket state per ${CLAUDE_PLUGIN_ROOT}/references/ticketing.md.
    No key or no plan is a fact to state in one line, not an error.
  3. gh pr view --json number,state,reviewDecision,mergeStateStatus and
    gh pr checks --required. Read them now; a status from before the last push is stale.
  4. git log --oneline origin/<default>..HEAD (at most 10) plus the uncommitted diff, boiled down to
    what changed and why.
  5. Agents or background tasks this session still runs (ListAgents where available). A running
    worker turns "nothing to do" into "wait for X".

Output

STANDUP — <KEY> · <branch> · PR #<n> <state> · CI <green|n red|pending> · tree <clean|n dirty>
ACHIEVED  <≤ 3 lines: what is now true — a merged PR, a green gate, a closed phase>
NEXT      <story key(s) and the command that starts them — or: nothing queued>
YOUR MOVE <one action with its command — or: nothing, because …>
  • The move is one item. With two candidates, name the one that unblocks the other.
  • More than three things owed means the story is not finished: say that instead of a list.
  • A missing source (no gh, no plan, no ticketing) is named as a gap, never filled by inference.
  • The session title should carry the ticket key; if it does not, print /rename <KEY> once.