Use case

AI Workflows for Weekly Status Reports

Name: Fleet
Author: Fleet

Status reports are the tax every team lead pays: an hour each Friday reconstructing what happened from memory, commit logs, and ticket queues. Skip a week and stakeholders fill the silence with assumptions. Delegate it and the report drifts into vagueness, because the person writing it wasn't in the work.

The information already exists — in merged PRs, closed tickets, decision docs, and planning files. What's missing is a reliable process that assembles it, checks it, and ships it on schedule without anyone remembering to.

How it works with an agent fleet

A scheduled Fleet workflow drafts the report from the repository itself — the planning docs, decision records, and notes you already commit — then routes it through review and a manager sign-off before it goes anywhere.

genflows:
  - name: weekly-status
    schedule: "0 16 * * 5"   # Fridays, 4pm
    steps:
      - {name: digest, prompt: "Summarize this week's shipped work, in-progress items, and blockers from the corpus.", corpus: ["docs/decisions/*.md", "planning/**/*.md"], kind: report, out: weekly-status.md}
      - {name: review, prompt: "Flag anything overstated, missing, or stale.", depends_on: [digest], kind: review, out: review.md}
      - {name: lead-signoff, depends_on: [digest, review], kind: approval, out: decision.md}
      - {name: announce, depends_on: [lead-signoff, digest], kind: notify, out: sent.md}

The cron schedule lives in the workflow, so the report happens whether or not anyone remembers. The lead gets an approval email Friday afternoon, reads the draft inline, edits the audit note if needed, and approves — the notify step then delivers it.

The fleet pattern

Schedule trigger → digest from corpus → review → approval gate → notify. The human stays the sender of record: nothing leaves without the sign-off, but the assembly and first-pass accuracy check are no longer their Friday afternoon.

Guardrails that matter here

The approval gate means the report is never auto-sent — the lead approves the exact text every week
Incremental rebuild: steps whose inputs didn't change skip on re-run, so an unchanged week re-uses the unchanged sections' work
Per-run audit trail — each week's report, review findings, and approval are versioned artifacts you can revisit

Who this is for

Team leads and managers who owe a recurring written update and already keep planning or decision docs in a repo. Works for engineering standup digests, exec summaries, and cross-team broadcasts.

Frequently asked questions

Where does the workflow get its information?

From the corpus globs you declare — markdown files in your repository. Teams that keep decision records, planning docs, or changelogs in git get good drafts immediately. The corpus is resolved with the same machinery as `fleet corpus resolve`, honoring .gitignore.

Can it post to Slack or email the team?

The notify step delivers to named recipients, and approval emails are built in. The report artifact is also a published, shareable document — most teams link it from wherever the team already reads updates.

Related use cases

AI Workflows for Board Reporting →AI Workflows for Recurring Data Reports →AI Workflows for Documentation After Every PR →

Run your first agent fleet

One binary. Five minutes. See every agent, coordinate every handoff, and keep a full audit trail of what your fleet did.

See how it works Install Fleet