The interface is localized. Skill descriptions and documents are machine-translated by AI into your language; code, commands, paths and technical terms stay in the original, and the author's document governs.
Explore skills
Development · SKILL.md

standup

standup

Quick project status report synthesized from state files, artifacts, and git activity. No agent dispatch (<2s).

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill documentation

GitHub

name: standup description: Quick project status report synthesized from state files, artifacts, and git activity. No agent dispatch (<2s).

AgenTeam Standup

Produce a fast, Linear-style project status report by reading the current run state, role artifacts, and git activity. This skill does not dispatch subagents -- the lead agent reads everything directly and synthesizes.

Process

1. Auto-Init Guard

Check for .agenteam/config.yaml, .agenteam.team/config.yaml, or legacy agenteam.yaml in the project root. If all are missing:

  • Create config dir: mkdir -p .agenteam
  • Copy the template: cp <plugin-dir>/templates/agenteam.yaml.template .agenteam/config.yaml
  • Set the team name to the project directory name
  • Generate agents: python3 <runtime>/agenteam_rt.py generate
  • Tell the user: "AgenTeam auto-initialized with default roles. Edit .agenteam/config.yaml to customize."

2. Gather State

Call the runtime to assemble run state, health indicator, and artifact paths:

python3 <runtime>/agenteam_rt.py standup

Capture the JSON output. Expected fields:

  • health -- on-track, at-risk, off-track, or no-active-run
  • run_id -- current run identifier (may be null if no active run)
  • task -- task description
  • stages -- list of stages with status, assigned role, and gate state
  • artifact_paths -- map of role name to artifact directory
  • governance.adoption -- optional summary of decisions, escalations, tripwire checks, gate rejections, and criteria overrides
  • output_path -- where to write the final report (e.g., docs/meetings/<timestamp>-standup.md)
  • Active attempt details -- thread_id, last heartbeat and heartbeat age, idle budget and wall budget remaining, retry count, and stop reason

3. Read Role Artifacts

For each role in artifact_paths, check what exists in that directory and produce a one-line summary:

RoleArtifact DirectoryLook For
researcherdocs/research/Research reports, dated findings
pmdocs/strategies/Strategy docs, roadmaps, specs
architectdocs/designs/Design docs, architecture decisions
devsrc/**, docs/plans/Source files changed, plan docs
qatests/**Test files, coverage reports
reviewer(read-only)Review comments in state files

For each role with artifacts:

  • List the most recent files (by modification time)
  • Summarize in one line: what was produced or changed

If no active run exists, scan artifacts and git history to build a best-effort summary of recent team activity.

4. Check Git Activity

Briefly inspect recent git state:

git log --oneline -10
git branch --list
git status --short

Summarize at the changeset level:

  • Recent commits (group by area, not raw list)
  • Active branches relevant to team work
  • Uncommitted changes (if any)

Do not include raw commit hashes or full diffs. Report outcomes, not activity.

5. Synthesize Report

Produce a Linear-style status report using this format:

Health: [ON TRACK | AT RISK | OFF TRACK]

## Completed
- [role] one-line summary of what's done

## In Progress
- [role] what's happening now

## Blocked
- [role] problem + next step + owner

## Decisions
- key decisions made (from design docs if available)

## Next
- what happens when current work completes

Rules for the report:

  • Health indicator comes first. Derive from the runtime JSON health field. If stages are blocked or gates rejected, it is AT RISK or OFF TRACK.
  • Completed lists roles whose stages are done, with a one-line summary of their artifact output.
  • In Progress lists roles with active stages, describing current work. Include thread_id, last heartbeat, idle budget, and wall budget for long-running work.
  • Blocked lists any role that is stuck. Every blocker must include three parts: the problem, the proposed next step, and the owner responsible for unblocking.
  • An interrupted role must include its stop reason and whether the persisted thread can be resumed. A stale heartbeat is an at-risk signal, not automatic authorization for fresh redispatch.
  • Decisions captures key architectural or strategic decisions found in governance.adoption, design docs, or strategy files. Include open follow-ups, escalations, tripwire blocks, gate rejections, and criteria overrides when the runtime JSON reports them. Omit this section if none are found.
  • Next describes what happens once current in-progress work completes (the next stage in the pipeline, or follow-up actions).
  • Omit any section that would be empty.

6. Write Report

Write the synthesized report to the output_path from the runtime JSON (typically docs/meetings/<timestamp>-standup.md):

mkdir -p "$(dirname "$output_path")"

Write the report content to that file.

7. Display to User

Show the full report to the user in the conversation. If there is an active run, include the run context header:

AgenTeam Standup: <team-name>
Run: <run-id> | Task: <short task description>

If no active run, use:

AgenTeam Standup: <team-name>
(No active run -- summarizing recent activity)

Runtime Path Resolution

Resolve the AgenTeam runtime:

  1. If running from the plugin directory: ./runtime/agenteam_rt.py
  2. If installed as a Codex plugin: <plugin-install-path>/runtime/agenteam_rt.py

Performance Target

This skill should complete in under 2 seconds. It reads local files and runs lightweight git commands -- no LLM subagent calls, no network requests, no web searches.

Saved skills