# Daily Summary

> Generate a comprehensive daily changelog for the current project — reads Issues.md, CHANGES.md, git history, and the conversation to produce a thorough report, then converts to Word

- Skill: `searayca/daily-summary` (Agent Skill)
- Install (CLI): `npx skillmds@latest add searayca/daily-summary`
- Raw SKILL.md: https://api.skillmd.com/api/skills/searayca/daily-summary/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: searayca (https://skillmd.com/u/searayca)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/searayca/daily-summary

---

# Daily Summary Generator

Generate a comprehensive daily changelog for any git project.

**CRITICAL: Git commits alone tell less than 50% of the story.**

---

## Step 1 — Run the Data Collectors

```bash
# Git + project data
python3 ~/.claude/tools/generate_daily_changelog.py

# TODAY's session — reads live /tmp task files (current session, not yet archived)
python3 ~/.claude/tools/session_agent_report.py --markdown

# HISTORICAL trend — reads persisted log from prior sessions
# (today's data appears here only after this session ends and Stop hook fires)
# Use --project to filter to THIS project only (get project name from git remote or directory name)
python3 ~/.claude/tools/r2_stats.py --week --markdown --project <current-project-name>
```

---

## Step 2 — Read Key Sources Before Writing (MANDATORY)

- ✅ **Session summary at the top of this conversation** — if the context was compacted, this summary contains hours of work not visible in git. **Always read this first.**
- ✅ **`docs/Issues.md`** — read the full file; every issue is a story about what went wrong and how it was solved
- ✅ **`CHANGES.md`** — read the full file; every user-requested change with its status
- ✅ **Full git commit details** — commit bodies, not just subject lines
- ✅ **The conversation** — look for debugging sessions, rejected approaches, and decisions that never made it into a commit

If unsure whether anything was missed, ask: *"Did we do anything else today I should include?"*

---

## Step 2b — Read Today's Time Blocks

Read `~/.claude/time/blocks/$(date +%Y-%m-%d).json` and filter to the
**current project** (git repo basename for the cwd, with `project_names.json`
variant map applied).

Also read matching entries from `~/.claude/time/siri_time.log` for today.

Add a **"Time Logged Today"** section to the changelog listing every matching
block with its start–end time, hours, and summary. Tag Siri-sourced rows
with `📞 siri`. Example:

```
## Time Logged Today

- 09:12–11:30  2.30h  Fixed login redirect loop; added session tests.
- 14:05–16:00  1.95h  ⚠ unattended — Refactored template partials.
- 📞 siri  0.75h  Voice config tuning (retroactive).

Total: 5.00h
```

If the blocks file is missing, skip this section silently — it just means
the tracker hasn't yet run for today.

---

## Step 3 — Generate the Changelog

Include ALL of these sections:

1. **Executive Summary** — 2-3 sentences covering the entire day
2. **Change Requests** — from CHANGES.md with status (✅ / ⚠️ / ❌)
3. **Issues Logged & Resolved** — problem → solution for each
4. **Technical Details** — HOW things were implemented, with code snippets
5. **Problems Encountered** — what failed, what was tried, what finally worked
6. **Testing Results** — results and coverage
7. **Metrics** — commits, files, lines, build times, repo size, etc.
7a. **Agent Usage (Today)** — paste the FULL raw table from `session_agent_report.py --markdown` verbatim. Top 10 by cost, session total, total tokens. If no task files were found, write "No agent task data available (tmp files cleared)" — do NOT omit the section.
7b. **Token Trend (Week)** — paste the FULL raw table from `r2_stats.py --week --markdown --project <name>` verbatim. This project only — do NOT include other projects' token usage. Note: today's row appears after session ends. **This section is MANDATORY — always include it even if session agent data is missing.**
8. **Architecture Decisions** — why certain approaches were chosen
9. **Outstanding Work / Tech Debt** — incomplete items
10. **Next Steps** — clear, actionable items

---

## Step 4 — Save, Convert, and Archive

1. Save to `docs/md/[PREFIX] CHANGELOG_YYYY-MM-DD_v1.0.md`
2. Run `/md_to_word --files <file> --open`
3. Run `python3 ~/.claude/tools/archive_old_docs.py`
4. Run `python3 ~/.claude/tools/archive_session.py` — snapshot current session to R2 and write token log.
   (The Stop hook will re-run this when the session eventually closes, overwriting with the complete final archive.)