Coding Day Recap

An Agent Skill that generates a visual HTML dashboard summarizing your daily AI coding sessions. It mines session data from CASS, reads sessions deeply to build an accurate narrative, and produces a self-contained dark-themed dashboard. Compatible with any coding agent that supports the Agent Skills standard — Claude Code, GitHub Copilot, OpenAI Codex, Cline, Roo Code, and more.

What it does

1. Gathers quantitative data — session counts, hourly activity distribution, workspace breakdown, agent breakdown
2. Reads sessions deeply — launches parallel subagents to read the START, MIDDLE, and END of each significant session
3. Builds an accurate narrative — distinguishes "coded" from "reviewed" from "committed", traces activity threads across sessions, checks git history for ground truth
4. Generates an HTML dashboard — dark-themed, self-contained, with a Day Map swimlane visualization, journal feed, workspace breakdown, and a hero stat

Prerequisites

• A coding agent that supports Agent Skills — Claude Code, GitHub Copilot, OpenAI Codex, Cline, Roo Code, Windsurf, or any compatible tool
• CASS — Coding Agent Session Search. Indexes sessions from Claude Code, Cursor, Codex, Gemini, and more
• Python 3 — required for the HTML generation script (ships with macOS; included with CASS on other platforms)

Platform

macOS, Linux, Windows — all scripts are Python, no platform-specific dependencies.

What is an Agent Skill?

An Agent Skill is a reusable prompt package that extends AI coding agents with specialized capabilities. Skills are defined by a SKILL.md file and follow an open standard adopted by 27+ tools. Each tool stores skills in its own directory — see the installation section for common locations.

Installation

1. Install CASS and index your sessions:

   pip install cass
   cass index

2. Clone this repo:

   git clone https://github.com/Gabko14/Coding-Day-Recap.git

3. Symlink to your agent's skills directory:

   Tool	Skills directory
   Claude Code	~/.claude/skills/
   OpenAI Codex	~/.codex/skills/
   Cline	~/.cline/skills/ (also reads ~/.claude/skills/)
   Roo Code	~/.roo/skills/
   GitHub Copilot	.github/skills/ (per-project)

   # Example for Claude Code (macOS / Linux)
   mkdir -p ~/.claude/skills
   ln -s /path/to/Coding-Day-Recap ~/.claude/skills/day-summary

   # Example for Claude Code (Windows)
   New-Item -ItemType Junction -Path "$HOME\.claude\skills\day-summary" -Target "C:\path\to\Coding-Day-Recap"

4. Invoke the skill (syntax varies by tool):

   /day-summary today
   

Usage

Invoke the skill with a date argument (exact syntax depends on your tool):

/day-summary today
/day-summary yesterday
/day-summary 2026-02-09

The skill will:

1. Index your CASS data
2. Launch parallel subagents to deeply read your sessions
3. Synthesize findings into a coherent timeline
4. Generate an HTML dashboard on your Desktop
5. Open it in your browser

Subagent Architecture

The skill launches a single subagent to read all pre-extracted session files, browser history, and git history. Using one reader instead of many produces a more consistent narrative — it sees activities spanning time blocks naturally without needing cross-block merging or conflict resolution.

The main agent then uses the subagent's findings to build the final timeline and generate HTML.

How it works

Phase 1: Quantitative Data

Runs CASS queries to gather session counts, hourly distribution, workspace/agent breakdowns.

Phase 2: Narrative Building

The critical phase. A single subagent reads all pre-extracted session data, browser history, and git history to build a complete picture of the day. Using one reader produces consistent results — it sees activities spanning time blocks naturally without needing cross-block merging.

Key accuracy rules:

• Check git history to distinguish "coded today" from "committed code written last week"
• Trace activities across sessions, not within — many activities span multiple sessions
• Name activities by what was done, never by session size or tools used

Phase 3: HTML Generation

Writes a data JSON file following the schema in scripts/generate_html.py, then generates a self-contained HTML dashboard with:

• Day Map — swimlane visualization showing activity bands across the day
• Journal Feed — detailed cards for each activity with descriptions and tags
• Workspace Breakdown — which repos got the most activity
• Hero Stat — the most striking number of the day

File Structure

Coding-Day-Recap/
├── SKILL.md                # Agent Skill definition
├── scripts/
│   ├── generate_html.py    # JSON → HTML generator
│   ├── pre_extract.py      # Batch CASS session extraction
│   └── browser_history.py  # Browser history extraction (Chrome/Edge)
├── assets/
│   └── template.html       # HTML/CSS template
├── examples/
│   └── screenshot.png      # Example dashboard
├── LICENSE
└── README.md

License

MIT