daily-recap
Skill2 starsChanged 6 months agoMITView on GitHub
Reconstruct a chronological timeline of a working day from external activity — Slack, GitHub, calendar, meeting notes (Granola), and Claude Code sessions. Produces a markdown file with stand-up block, timeline, and stats. Use when the user says "recap", "/daily-recap", "what did I do today", "yesterday recap", or asks for a retrospective of a day's work.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- jefrykurniaone/grill-to-wavestheir own version, first seen Nov 15, 2026
- CodeAlive-AI/ceo-ai-ostheir own version, first seen May 18, 2026
- Layneformalized225/ai-cofoundertheir own version, first seen Mar 22, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- CodeAlive-AI/ceo-ai-os★ 9same name
- jefrykurniaone/grill-to-waves★ 9same name
- Layneformalized225/ai-cofounder★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository20 other files it gives its agents.
- 1-Projects/AGENTS.mdAGENTS.md
- 2-Areas/AGENTS.mdAGENTS.md
- 3-Resources/AGENTS.mdAGENTS.md
- 4-Archive/AGENTS.mdAGENTS.md
- _admin/AGENTS.mdAGENTS.md
- _agent/AGENTS.mdAGENTS.md
- Books/AGENTS.mdAGENTS.md
- Content/AGENTS.mdAGENTS.md
- Dashboard/AGENTS.mdAGENTS.md
- Goals/AGENTS.mdAGENTS.md
- Granola/AGENTS.mdAGENTS.md
- Inbox/AGENTS.mdAGENTS.md
- Journal/AGENTS.mdAGENTS.md
- People/AGENTS.mdAGENTS.md
- Recipes/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- add-moduleSkill
- close-daySkill
- ingest-meetingSkill
- todaySkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Use the argument if provided, otherwise today.
- Distinctive: line 20
Never guess the day-of-week — always compute it.
- Distinctive: line 61
Use those timestamps as the source of truth.
- Distinctive: line 141
No meeting-by-meeting detail or Slack chatter — anything that wouldn't belong in a standup gets dropped.
- Very specific: line 145
Write to _agent/outputs/recaps/<YYYY-MM-DD> - <DayOfWeek> - Recap.md (e.g., 2026-05-06 - Wednesday - Recap.md).
- Distinctive: line 147
If the file exists, read it first, merge in new entries, and preserve any manual edits.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Daily Recap — Retrospective Timeline
Reconstructs what happened during a working day from external sources. Pairs with close-day (which processes the user's own daily-note writing).
Date
Use the argument if provided, otherwise today.
TARGET_DATE="${ARG:-$(date +%F)}"
DAY_NAME=$(date -d "$TARGET_DATE" +%A) # macOS: date -j -f "%Y-%m-%d" "$TARGET_DATE" +%ANever guess the day-of-week — always compute it.
Conventions
- Vault root: operate from CWD. The skill auto-detects the Claude project directory matching the current path.
- Output:
_agent/outputs/recaps/<YYYY-MM-DD> - <DayOfWeek> - Recap.md. The directory_agent/outputs/exists; createrecaps/on first run. - Timezone: defaults to the user's local TZ from the shell. Override with
TZenv var. The "working day" boundary defaults to 03:00 local time on the target date through 03:00 local time the next day, capturing late-night sessions.
Prerequisites
Required:
- Google Calendar MCP — for events.
ghCLI authenticated — for GitHub events / PR enrichment. Skip GitHub steps ifgh auth statusfails.
Optional (skill degrades gracefully if absent):
- Slack MCP — at session start, scan available tools for
mcp__*[Ss]lack*(e.g.,search_messages,read_channel). If none, skip Slack steps and note in output. - Granola MCP (
mcp__*Granola*) — for meeting notes / actual durations. - Claude Code session logs at
~/.claude/projects/<slugified-vault-path>/*.jsonl— used for "what was worked on" attribution.
Step 1 — Gather data (parallel)
Run these concurrently:
- Slack messages from me — search
from:me on:<date>, sorted ascending, paginate. Also searchfrom:me on:<next-date>and keep only messages before 03:00 local time. - Slack messages to me — search
to:me on:<date>(same logic). - Calendar events — list events for
<date>in the local TZ. - GitHub activity —
.claude/skills/daily-recap/github-events.sh <date>. The script honorsTZ,EXCLUDE_REPOS(comma-separated), andWORKDAY_START_HOURenv vars. Output:HH:MM | EventType | repo | action=... | ref=... | pr=... | title=... | commits=[...] - Granola meetings —
mcp__*Granola*list_meetingsfor the target date. Use Granola times when both calendar and Granola exist (Granola has actual durations). - Claude Code sessions —
.claude/skills/daily-recap/claude-sessions.sh <date>for session summaries; add--promptsfor substantive user prompts. The script auto-derives the project scope from$PWD.
Slack timestamp accuracy
Slack search results return opaque message IDs, not real timestamps. To get accurate times:
- Collect the unique channel/DM IDs that had activity from the search results.
- For each active channel, use
slack_read_channel(or equivalent) witholdest/latestset to the target date's Unix-timestamp range. - Use those timestamps as the source of truth.
Step 2 — Enrich GitHub
For each unique PR number found, batch-fetch titles and diff sizes:
.claude/skills/daily-recap/pr-info.sh <owner/repo> <pr> [<owner/repo> <pr> ...]Output: owner/repo#<pr> | <title> | +<additions>/-<deletions>. Use diff size to estimate review time (small <100 ≈ 5 min, medium 100–500 ≈ 15 min, large >500 ≈ 30 min). Batch into one invocation to avoid per-call permission prompts.
Step 3 — Build the timeline
Combine sources into one chronological timeline. For each entry:
- Time (local TZ)
- Source (Slack channel/DM, Calendar, GitHub repo, Granola, Claude project)
- What happened — concise
Group related Slack messages into thread-level entries (one entry per thread, not per message). Merge corroborating signals (a GitHub push + Claude session for the same project = one entry).
For Claude Code sessions, one entry per distinct working window:
- Time range (from first/last substantive prompt in the window)
- Project (short name, not the slug)
- What was worked on (summarized from prompts, not a prompt-by-prompt dump)
Scope: work, not personal chatter
Filter out personal/social content even if it flowed through Slack/calendar (after-hours plans, restaurant logistics, etc.). Keep "company event happened" at calendar-level only. Mixed work+personal threads → summarize the work portion only.
Step 4 — Summary
End with a Summary section. Stats first, then takeaways.
Stats
Each stat on its own line; bulleted detail beneath where applicable:
- PRs Merged — count + bullets (repo, PR number, title)
- PRs Opened — count + bullets
- PRs Reviewed — count + bullets
- Commits — count, e.g., "6 commits across 2 repos" (excluding
EXCLUDE_REPOS) - Meetings — count + total duration + bulleted list (time, title)
- Code Review — estimated minutes from PR diff sizes (~5/15/30 by size)
- Slack — channels/DMs active in
- Lunch — detect 30–60 min activity gap typically 12:00–13:30; show range or "No obvious lunch break"
- Focus Time — blocks of 30+ min with no Slack and no meetings, work-hours only (infer end-of-workday from activity pattern; fall back to 18:00 if no signal). Exclude lunch. Sum and list specific blocks. If a block overlaps a Claude session, annotate the project worked on.
- Longest Focus Block — the single longest uninterrupted focus block.
- Evening Work (optional, informational) — substantive post-dinner work in one line; not counted as Focus Time.
- Claude Sessions — count of distinct sessions with substantive activity + projects touched.
Takeaways
2–3 sentences covering key discussions/decisions and main themes of the day.
Step 5 — Stand-Up Block
Generate a stand-up-ready block in Daily Notes/-style format. Place it first in the output file (before Timeline and Summary) — it's the most-referenced part.
- Yesterday = accomplishments from the previous working day (target date − 1). If target is Monday, label as Friday instead. To fill: read prior recap at
_agent/outputs/recaps/<prev>.md; otherwise leave a single empty bullet. - Today = accomplishments on the target date itself, drawn from this recap's takeaways + key timeline threads.
Format:
# Stand Up
**Yesterday**
- <terse bullet>
**Today**
- <terse bullet>Rules:
- Terse, action-phrased, no trailing punctuation. Match the style of any existing stand-ups in the vault.
- 1 to ~10 bullets. Group related work where natural ("Reviews — A, B, C"); don't force compression if the day had distinct threads.
- PR / TASK references must include a short subject, not just IDs.
PR#8343 rocket-engine expression library✓ —PR#8343✗. - No meeting-by-meeting detail or Slack chatter — anything that wouldn't belong in a standup gets dropped.
Step 6 — Save
Write to _agent/outputs/recaps/<YYYY-MM-DD> - <DayOfWeek> - Recap.md (e.g., 2026-05-06 - Wednesday - Recap.md). Create _agent/outputs/recaps/ if missing.
If the file exists, read it first, merge in new entries, and preserve any manual edits. Don't duplicate.
Tag the file with source: agent in frontmatter per _agent/AGENTS.md.
Output format
Markdown with ### for time blocks, --- between sections, bold for channel/repo names. Order in the file:
# <YYYY-MM-DD> - <DayOfWeek> - Recapheader# Stand Upblock## Timeline(Morning / Lunch / Afternoon / Evening)## Summary(Stats + Takeaways)
Caveats
- Claude session file span ≠ working window. A session file can stretch across hours while real activity is clustered. Use
--promptstimestamps for the truth. - Cross-day sessions are handled by the working-day window (03:00 → 03:00 next day).
- Sibling project dirs like
claude-b,claude-investigateare ephemeral scratch workspaces — name them by what was worked on (from prompts), not the dir name.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

