Weekly metrics report
Monday morning report of last week's shipping and product numbers, saved to reports/ with week-over-week change.
- Installs
- 0
- Runs (30d)
- 0
- Active builders
- 0
- Trend
- —
Uses
Tools: cli:git, composio:github, composio:posthog
Triggers: Mondays at 8:00 AM
Definition (YAML)
id: weekly-metrics-report
name: Weekly metrics report
category: productivity
tagline: Monday morning report of last week's shipping and product numbers, saved to reports/ with week-over-week change.
steps:
- "Counts last week's commits, pull requests and releases"
- "Adds product numbers from connected sources"
- "Saves a report to reports/ with week-over-week change"
requires:
- kind: cli
id: git
why: Counts commits, contributors and files changed.
- kind: composio
id: github
optional: true
why: Adds merged pull requests, review time and issues opened and closed.
- kind: composio
id: posthog
optional: true
why: Adds active users and your top events.
triggers:
- kind: schedule
cron: "0 8 * * 1"
install:
agentMode: Build
speed: balanced
runMode: single
autoApproveTools: true
maxRunMinutes: 20
replyStartsWith:
- "reports/weekly-"
prompt: |
Write the weekly metrics report for {{repo}} ({{target.path}}) for the 7 days before {{date}}, and compare each number with the 7 days before that.
Collect what is available:
- **Shipping (git)**: commits on the default branch, distinct authors, files changed and lines added/removed (`git log --since=... --until=... --shortstat --no-merges`), and the 3 largest changes by subject.
- **Pull requests and issues (GitHub via the `integrations` MCP server, if connected)**: merged pull requests, median hours from open to merge, issues opened vs closed.
- **Product (PostHog via the `integrations` MCP server, if connected)**: weekly active users and the 5 most frequent events.
Use YOLOSESH_SEARCH_TOOLS to find tools and YOLOSESH_MULTI_EXECUTE_TOOL to call them. Never start a connection flow. A source that is not connected gets one table row reading "not connected — connect <name> in Automations".
Compute every number with a command or a tool result, never estimate. Show change as +n / −n and percent.
Save the report to `reports/weekly-{{date}}.md` (create the folder if needed) and change nothing else.
Report format:
# Weekly metrics · {{repo}} · week ending {{date}}
| Metric | This week | Last week | Change |
Then **Highlights** (3 bullets) and **Watch** (anything that dropped more than 20%).
Reply with only this, with nothing before or after it: the line `reports/weekly-{{date}}.md`, then the Highlights. Missing sources and caveats belong in the report, not the reply.Agent script
Write the weekly metrics report for {{repo}} ({{target.path}}) for the 7 days before {{date}}, and compare each number with the 7 days before that.
Collect what is available:
- **Shipping (git)**: commits on the default branch, distinct authors, files changed and lines added/removed (`git log --since=... --until=... --shortstat --no-merges`), and the 3 largest changes by subject.
- **Pull requests and issues (GitHub via the `integrations` MCP server, if connected)**: merged pull requests, median hours from open to merge, issues opened vs closed.
- **Product (PostHog via the `integrations` MCP server, if connected)**: weekly active users and the 5 most frequent events.
Use YOLOSESH_SEARCH_TOOLS to find tools and YOLOSESH_MULTI_EXECUTE_TOOL to call them. Never start a connection flow. A source that is not connected gets one table row reading "not connected — connect <name> in Automations".
Compute every number with a command or a tool result, never estimate. Show change as +n / −n and percent.
Save the report to `reports/weekly-{{date}}.md` (create the folder if needed) and change nothing else.
Report format:
# Weekly metrics · {{repo}} · week ending {{date}}
| Metric | This week | Last week | Change |
Then **Highlights** (3 bullets) and **Watch** (anything that dropped more than 20%).
Reply with only this, with nothing before or after it: the line `reports/weekly-{{date}}.md`, then the Highlights. Missing sources and caveats belong in the report, not the reply.