| name | agents-standup |
| description | (beta) Run the daily AI-agent standup with yeaboi: what the user's coding agents did — local sessions worked (with estimated cost) plus agent-authored commits and PRs found in GitHub/Azure DevOps. Use when the user asks what their agents did, wants an agent standup/digest, or a daily agent activity summary. |
Agent standup workflows with yeaboi
Beta. The Agents modes are in beta — detection is a lower bound: agents
that leave no marker are invisible. Never present absence of evidence as
idleness, and present costs as estimates.
-
Run the digest with agents_standup_run. Defaults cover everything
since the previous working day (a Monday run reaches back to Friday) and
scan both trackers best-effort; narrow with:
days — explicit look-back window (0 = previous-working-day default).tracker_sources — ["github"], ["azdo"], or [] for local-only.github_owners / azdo_projects — scope the tracker scan.deliver — posts to the configured Slack webhook. Leave it false and
present the digest yourself; only set true if the user explicitly asks.
-
Present it like a standup: lead with the narrative, then highlights
(shipped work), in_flight (open agent PRs), and attention_items (what
needs a human). The session_summaries and repo_activity rows are the
evidence behind every line.
-
Surface coverage_notes and warnings — a tracker that couldn't be
scanned (missing token) is a visible gap, not an empty day.
-
Compare days with agents_standup_history (newest first) instead of
recomputing.
-
Privacy. Only aggregates and tracker metadata are read; session
transcripts are never copied or uploaded. Exports auto-save under
~/.yeaboi/exports/agentwatch/standup/.
Error handling
Every tool returns {ok, llm_mode, warnings, data}. If ok is false, relay
error.message and its hint; don't retry blindly. llm_mode: "fallback"
means no LLM was reachable — the evidence rows are still real, only the
narrative prose fell back to deterministic lines.
