nsheaps/daily-report

Daily Organization Report

Generate a comprehensive daily activity report for a GitHub organization. The report covers all repositories with activity in the specified time window, including commits, pull requests, branches, issue changes, and force-push tracking.

Inputs

The following environment variables or prompt context control behavior:

| Variable | Default | Description | | ---------------------- | ------------------------ | ---------------------------------------------------------------------- | | REPORT_TIME_SCOPE | 1 day | Free-form time scope, e.g. "1 day", "3 days", "1 week" | | REPORT_SUBJECT_SCOPE | nsheaps | Free-form subject scope, e.g. org name, specific repo, or email filter | | REPORT_DATE | Yesterday (Eastern Time) | The date the report covers, format YYYY-MM-DD |

Data Gathering

Use gh CLI with the available GH_TOKEN for all GitHub API calls. The token must have org-level read access to repositories, issues, and pull requests.

Step 1: Identify Active Repositories

# List all repos in the org
gh repo list "${ORG}" --limit 500 --json nameWithOwner,pushedAt -q '.[] | select(.pushedAt >= "'${SINCE_DATE}'") | .nameWithOwner'

If the subject scope names a specific repository, only report on that repo. If the subject scope includes an email filter, gather from all repos but filter commits by that author email.

Step 2: Gather Commits Per Repository

For each active repository, gather commits across ALL branches:

# All commits in the time window across all branches
gh api "repos/${OWNER}/${REPO}/commits" \
  --paginate \
  -q '.[] | {sha: .sha, message: .commit.message, author: .commit.author.name, email: .commit.author.email, date: .commit.author.date, branch: "N/A"}' \
  -f since="${SINCE_ISO}" -f until="${UNTIL_ISO}"

For more granular per-branch data, also list branches and check commits per branch:

# List all branches
gh api "repos/${OWNER}/${REPO}/branches" --paginate -q '.[].name'

# Commits on a specific branch
gh api "repos/${OWNER}/${REPO}/commits?sha=${BRANCH}&since=${SINCE_ISO}&until=${UNTIL_ISO}" --paginate

Step 3: Detect Force Pushes and Trace Previous Commit Trees

Force pushes rewrite history. To detect them, check the audit log and reflog events:

# Check for push events that were force pushes (org audit log)
gh api "/orgs/${ORG}/audit-log" \
  -f phrase="action:git.push created:>=${SINCE_DATE}" \
  -f include=git \
  --paginate \
  -q '.[] | select(.force_push == true) | {repo: .repo, actor: .actor, before: .before, after: .after, timestamp: .created_at}'

If the audit log is not available (requires GitHub Enterprise or org admin), use the events API as a fallback:

# Push events for a repo (includes force push info)
gh api "repos/${OWNER}/${REPO}/events" \
  --paginate \
  -q '.[] | select(.type == "PushEvent") | {ref: .payload.ref, size: .payload.size, before: .payload.before, head: .payload.head, forced: .payload.forced, actor: .actor.login, created_at: .created_at}'

When a force push is detected:

1. Record the before SHA (the previous HEAD)
2. Record the after SHA (the new HEAD)
3. Use gh api repos/${OWNER}/${REPO}/compare/${BEFORE}...${AFTER} to show what changed
4. Try to list the commits that were on the old tree: gh api repos/${OWNER}/${REPO}/commits?sha=${BEFORE} (these may be garbage-collected)
5. Include both the old and new commit trees in the report under a "Force Push History" subsection

Step 4: Gather Pull Requests

# PRs updated in the time window (all states)
gh pr list -R "${OWNER}/${REPO}" \
  --state all \
  --json number,title,state,author,createdAt,updatedAt,mergedAt,url,headRefName,baseRefName,additions,deletions \
  -L 500

Filter results to PRs with updatedAt within the time window.

Step 5: Gather Branch Activity

# All branches with their last commit info
gh api "repos/${OWNER}/${REPO}/branches" --paginate \
  -q '.[] | {name: .name, sha: .commit.sha, protected: .protected}'

# Recently created or deleted branches (from events)
gh api "repos/${OWNER}/${REPO}/events" \
  --paginate \
  -q '.[] | select(.type == "CreateEvent" or .type == "DeleteEvent") | select(.created_at >= "'${SINCE_ISO}'") | {type: .type, ref: .payload.ref, ref_type: .payload.ref_type, actor: .actor.login}'

Step 6: Gather Issue Changes

# Issues updated in the time window
gh issue list -R "${OWNER}/${REPO}" \
  --state all \
  --json number,title,state,stateReason,author,createdAt,updatedAt,url,labels \
  -L 500

Filter to issues with updatedAt within the time window. Also track:

• Newly opened issues
• Closed issues (with state reason: completed vs not_planned)
• Label changes
• Assignee changes

For issue timeline events:

gh api "repos/${OWNER}/${REPO}/issues/${ISSUE_NUMBER}/timeline" --paginate \
  -q '.[] | select(.created_at >= "'${SINCE_ISO}'")'

Step 7: Gather CI Workflow Failures (for Attention Required section)

# Recent workflow runs with failures
gh api "repos/${OWNER}/${REPO}/actions/runs" \
  -f created=">=${SINCE_DATE}" \
  -f status="failure" \
  --paginate \
  -q '.workflow_runs[] | {name: .name, branch: .head_branch, conclusion: .conclusion, url: .html_url, created_at: .created_at, run_number: .run_number}'

Look for patterns of repeated failures on the same branch/PR, which may indicate agent struggles. Cross-reference with PR and commit data from earlier steps.

Report Format

The output must be valid GitHub-flavored Markdown suitable for posting as a GitHub issue body.

Title

YYYY-MM-DD Daily Report

Where YYYY-MM-DD is the report date (the day being reported on).

Structure

The report is a single Markdown document. The template below shows the required structure. Sections marked with {placeholders} should be filled with actual data.

Header

# YYYY-MM-DD Daily Report

> **Scope**: {subject_scope} | **Period**: {time_scope} | **Generated**: {timestamp} ET

Attention Required

This section appears immediately after the header, before any other content. It draws the reader's attention to items that need human intervention or investigation.

Analyze the gathered data to identify two categories of items:

1. Agent Failures & Issues — Detect patterns that suggest automated agents struggled or failed during the reporting period:

• PRs with many force pushes on the same branch in a short time (suggesting repeated failed attempts)
• Commits with struggle-indicator messages ("retry", "attempt", "revert", "try again", "wip", "broken") appearing in rapid succession on the same branch. Exclude conventional commit fix: or fix(scope): prefixes — those indicate intentional bug fixes, not agent struggles. Only flag "fix" when 3+ fix-like commits appear within 1 hour on the same branch (suggesting iterative failed attempts rather than normal development)
• PRs that were opened and closed without merging multiple times
• CI workflow runs that failed repeatedly on the same PR/branch
• Branches with an unusually high number of commits relative to the diff size (churn)
• Any commit messages explicitly mentioning errors, failures, or workarounds

2. Action Items Requiring Human Involvement — Identify items from commits, PRs, and issues that explicitly call out the need for human action:

• References to secrets that need to be created or rotated
• Comments or PR descriptions mentioning "needs review", "need approval", or similar
• Issues labeled with status:blocked or status:needs-review
• References to permissions, access tokens, or credentials that need setup
• Deployment or infrastructure changes that require manual steps
• Any TODO/FIXME comments added in the reporting period that reference manual action

Format this section as a blockquote callout with warning emojis:

> ⚠️🚨 **ATTENTION REQUIRED** 🚨⚠️
>
> ### 🔴 Agent Failures & Repeated Iterations
>
> {List each detected issue with repo, branch/PR link, and brief description}
> {If none detected: "No agent failure patterns detected."}
>
> ### 🟡 Items Needing Your Involvement
>
> {List each item with repo, issue/PR/commit link, and what action is needed}
> {If none detected: "No items requiring manual intervention."}

Top Repos by Commits

A horizontal bar chart showing the total number of commits per repository during the reporting period. This gives a quick visual overview of where the most activity occurred.

• Use a mermaid xychart-beta chart with horizontal layout
• X-axis: repository short names (without the org prefix), sorted by commit count descending
• Y-axis: total commit count — ALWAYS start at 0. You MUST use the syntax 0 --> <max> where <max> is the highest bar value rounded up to a clean number. Never omit the 0 --> prefix or the axis will not start at zero.
• One bar per repository
• Use all commits gathered in Step 2 (across all branches)
• Only include repos that had at least 1 commit

Design note: This chart intentionally replaces the previous time-of-day stacked bar chart. The temporal dimension (when activity occurred) is traded for a clearer at-a-glance view of where activity concentrated. Time-of-day patterns can still be inferred from the commit tables in the Repository Activity section below.

Example chart (actual values will differ):

xychart-beta horizontal
    title "Commits by Repository (YYYY-MM-DD)"
    x-axis ["repo-a", "repo-b", "repo-c"]
    y-axis "Commits" 0 --> 20
    bar [18, 9, 5]

Remaining Report Sections

After the attention callout and chart, include these sections in order:

## Executive Summary

One paragraph summarizing the day's activity across the org: total commits, PRs opened/merged/closed, issues opened/closed, repos with activity, notable force pushes.

## Repository Activity

### {owner}/{repo-name}

For each active repo (sorted alphabetically), include:

• Commits ({count}) — table with SHA (linked), Author, Branch, Message, Time (ET)
• Pull Requests ({count}) — table with #, Title, State, Author, Branch, +/-, Updated
• Branch Activity — table with Branch, Event, Actor, Time
• Issue Changes ({count}) — table with #, Title, Change, State, Labels
• Force Push History — only if force pushes detected; table with Branch, Actor, Time, Before/After SHAs (linked), Commits Lost/Added. Use <details> for old commit tree.

Separate each repo with ---.

## Cross-Repo Summary

• By Author — table with Author, Commits, PRs, Issues

• Activity Timeline — chronological list of significant events across all repos (ET)

  Methodology Notes

• Time window: {SINCE_ISO} to {UNTIL_ISO}

• Repositories scanned: {count}

• Repositories with activity: {count}

• Force push detection: {method used - audit log or events API}

• Any data gaps or API limitations encountered

Formatting Rules

1. All references must be clickable links: Every SHA, issue number, PR number, and branch name that can link to GitHub must be a markdown link
2. Times in Eastern Time: Convert all UTC timestamps to US Eastern Time for display
3. Use collapsible sections (<details>) for verbose data like force-push old commit trees
4. Sort repos alphabetically within the report
5. Sort commits chronologically (oldest first) within each repo section
6. Include zero-activity note: If a repo had no activity, do not include it
7. Differentiate issue close reasons: Show "closed (completed)" vs "closed (not_planned)"

Error Handling

• If the org audit log is inaccessible, note this in Methodology and fall back to events API
• If a repo returns 404 (private/deleted), skip it and note in Methodology
• If rate limiting occurs, note partial data in Methodology
• If commits are garbage-collected after force push, note "commits no longer available"

Output

The final report content should be posted as a GitHub issue in the triggering repository with:

• Title: YYYY-MM-DD Daily Report
• Label: daily-report
• The full markdown report as the issue body