daily-report — Claude Code Skill

A Claude Code skill that scans your ~/.claude/projects/ session logs and generates a structured daily work report with timeline, deliverables, signals, and optional Scrum standup preparation.

What it does

1. Scans ~/.claude/projects/**/*.jsonl for today's Claude Code sessions
2. Launches 3 parallel Haiku sub-agents to analyze:
   • Timeline — chronological workstream summary
   • Deliverables — files written, reports generated, builds run
   • Signals — blockers, repeated issues, unfinished items
3. Renders a paired report: /tmp/daily-report-YYYY-MM-DD-report.md + .html
4. Opens the HTML in Chrome automatically
5. Optional Scrum standup — add "standup" / "站会" / "晨会" to get the 3-question format

No Python scripts. No hard-coded dependencies. Uses Claude's native tools only.

---

Installation

macOS

Step 1 — Install the skill

# Create the global skills directory if it doesn't exist
mkdir -p ~/.claude/skills

# Clone or copy the skill
cp -r daily-report ~/.claude/skills/daily-report

Step 2 — Verify before committing (test in /tmp)

# Create a clean test workspace
mkdir -p /tmp/dr-test-workspace
cd /tmp/dr-test-workspace

# Ask Claude to explain the skill (should NOT say "Unknown skill")
claude --model claude-sonnet-4-6 -p '/daily-report do not run just explain what this would do?'

✅ Expected output: A full explanation of the 4-step workflow
❌ Bad output: Unknown skill: daily-report

Step 3 — Run it for real

cd /tmp/dr-test-workspace

# Basic daily report
claude -p '/daily-report'

# With Scrum standup section
claude -p '/daily-report standup'

---

Windows

Requires: Claude Code CLI + Git Bash or PowerShell

Step 1 — Install the skill

# In PowerShell or Git Bash
$skillsDir = "$env:USERPROFILE\.claude\skills"
New-Item -ItemType Directory -Force -Path $skillsDir

# Copy the skill folder
Copy-Item -Recurse daily-report "$skillsDir\daily-report"

Or with Git Bash:

mkdir -p ~/.claude/skills
cp -r daily-report ~/.claude/skills/daily-report

Step 2 — Verify before committing (test in temp folder)

Open a new terminal and run from a neutral directory:

# PowerShell
cd $env:TEMP
claude --model claude-sonnet-4-6 -p '/daily-report do not run just explain what this would do?'

Or Git Bash:

cd /tmp
claude --model claude-sonnet-4-6 -p '/daily-report do not run just explain what this would do?'

✅ Expected: Full 4-step explanation
❌ Bad: Unknown skill: daily-report

Step 3 — Run it for real

cd $env:TEMP
claude -p '/daily-report'

# With Scrum standup
claude -p '/daily-report standup'

Note (Windows): The skill auto-opens the HTML report with open -a "Google Chrome" which is macOS-only. On Windows, open the report manually:

Start-Process "C:\Users\$env:USERNAME\AppData\Local\Temp\daily-report-YYYY-MM-DD-report.html"

---

Verification Checklist

Run these checks after installation:

# 1. Skill is recognized (no "Unknown skill" error)
cd /tmp && claude -p '/daily-report do not run just explain what this would do?'

# 2. Skill files are in the right place
ls ~/.claude/skills/daily-report/
# Should show: SKILL.md  references/  scripts/  assets/

# 3. SKILL.md is readable
cat ~/.claude/skills/daily-report/SKILL.md | head -5

---

Usage

Command	What it does
/daily-report	Today's report
/daily-report 2026-04-12	Report for a specific date
/daily-report standup	Today's report + Scrum 三问
/daily-report standup 2026-04-12	Specific date + standup

---

Output Files

All output goes to /tmp/ (or %TEMP% on Windows):

/tmp/daily-report-YYYY-MM-DD-report.html   ← main report (auto-opens in Chrome)
/tmp/daily-report-YYYY-MM-DD-report.md     ← Markdown version
/tmp/daily-report-YYYY-MM-DD-timeline.md   ← Haiku agent output
/tmp/daily-report-YYYY-MM-DD-deliverables.md
/tmp/daily-report-YYYY-MM-DD-signals.md
/tmp/daily-report-YYYY-MM-DD.data.jsonl    ← raw scan data

---

Requirements

• Claude Code CLI installed and authenticated
• At least one Claude Code session logged today in ~/.claude/projects/
• Google Chrome (for auto-open on macOS; manual open on Windows)

---

Skill Structure

daily-report/
├── SKILL.md                        ← skill definition (no scripts)
└── references/
    └── agent-team-contract.md      ← 3-agent analysis contract

No bundled Python scripts. All scanning and rendering is done with Claude's native Bash/Read/Write/Glob tools.

---

License

MIT