Daily Summary Skill

Produce a short German bullet list of today's work across all Claude Code sessions, grouped by project or repository.

Step 1: Get today's date

python3 -c "from datetime import date; print(date.today().isoformat())"

Step 2: Load the memory tools

Use ToolSearch to load the required tool schemas in one call:

ToolSearch with query select:mcp__plugin_claude-mem_mcp-search__search,mcp__plugin_claude-mem_mcp-search__get_observations

Step 3: Fetch today's observations

Call mcp__plugin_claude-mem_mcp-search__search with dateStart and dateEnd both set to today's ISO date (e.g. 2026-06-24), limit 100, orderBy date_asc. The tool returns a list of observations for the day.

Focus on actionable observation types:

Session entries (prefix S) provide the headline title — use these to name the project if content is ambiguous.

Step 4: Group by project

Map each session/observation to a project or repository. Common signals:

• Repo or package name in the session title (e.g., app-startpage, skills, translations-app)
• File paths or package names mentioned in observation text
• Branch names or PR references

If a session spans multiple repos, split into separate bullets per repo.

Step 5: Produce the German summary

Output directly in the conversation — no file needed.

Format

## Tagesrückblick — DD. Monat YYYY

- [x] **Projektname**: Ein Satz im Perfekt, was heute erledigt wurde.
- [x] **Projektname**: Ein Satz im Perfekt, was heute erledigt wurde.

Rules

• One checked off bullet per project — dense, not exhaustive
• German Perfekt tense (wurde X implementiert, wurden Y Fehler behoben)
• Bold the project/repo name
• Date in German long form: 24. Juni 2026
• Omit projects where only trivial formatting or config changes happened — unless that was the only work
• If no sessions found: output Heute wurden keine Claude-Code-Sitzungen aufgezeichnet.
• Max one sentence per bullet — if more happened, pick the most significant item