| name | daily-brief |
| description | Generate and manage daily project briefs. Reads from and writes to open-brain as the system of record (type=daily_brief). Disk write is opt-in via --persist-disk. Use for daily summaries of git activity, bead progress, and open-brain entries across tracked projects.
|
| triggers | ["daily brief","daily briefing","tagesbericht","was hatte ich gestern gemacht","daily-brief","brief generieren","/daily-brief"] |
| requires_standards | ["english-only"] |
daily-brief
Generate a daily brief for one or all tracked projects, covering git commits,
bead activity, and memory entries for a given date range.
Quick Start
All CLI forms delegate to scripts/orchestrate-brief.py:
python3 scripts/orchestrate-brief.py
See the Orchestration section below for all supported CLI args.
Config
Config lives at ~/.claude/daily-brief.yml. Bootstrap it with:
python3 core/skills/daily-brief/scripts/config.py load
This creates the file with four default projects (claude-code-plugins, mira, polaris,
open-brain) if it does not already exist.
Config Schema
projects:
- name: claude-code-plugins
path: /Users/malte/code/claude-code-plugins
slug: claude-code-plugins
beads: true
docs_dir: docs
defaults:
since: yesterday
format: markdown
detailed: false
language: de
timezone: Europe/Berlin
Storage Layout (v1.5)
open-brain is the primary system-of-record. Disk is opt-in via --persist-disk.
open-brain (primary):
type=daily_brief, project=<slug>, session_ref=daily-brief-{slug}-{date}
metadata: {do_not_compact: true, schema_version: "1.5"}
~/.claude/projects/<slug>/daily-briefs/ (opt-in, --persist-disk only):
2026-04-24.md
2026-04-23.md
...
Migration: Run python3 scripts/migrate-disk-briefs-to-open-brain.py --apply
once to import existing disk briefs into open-brain. See ADR-0002 for the full
migration playbook.
Orchestration (scripts/orchestrate-brief.py)
The orchestrate-brief.py script is the main entry point. It handles all CLI
args, backfill logic, and open-brain persistence.
Orchestration flow (v1.5):
- Parse CLI args
- Load config, resolve project(s)
- For each (project, date): check open-brain first → if found, return cached content
- Fallback: check disk (for offline / OB-unavailable scenarios)
- If neither: call
render-brief.py (with --no-persist unless --persist-disk)
- Persist each new brief to open-brain (hard error on failure)
- Emit aggregated markdown
Backfill rule: Re-running the same args is a no-op. open-brain is checked
first; if the brief is already there, it is returned without re-rendering.
Open-brain persistence (v1.5 — hard error): Each new brief is saved as
type=daily_brief, project=<slug>, session_ref=daily-brief-{slug}-YYYY-MM-DD,
do_not_compact=true. Failure to write raises — OB credentials are required.
CLI args:
| Invocation | Behaviour |
|---|
python3 scripts/orchestrate-brief.py | All config projects, since=yesterday |
python3 scripts/orchestrate-brief.py <name> | Single project, since=yesterday |
python3 scripts/orchestrate-brief.py --since=Nd | Last N days (2d, 7d, etc.) |
python3 scripts/orchestrate-brief.py --date=YYYY-MM-DD | Specific date |
python3 scripts/orchestrate-brief.py --range=YYYY-MM-DD..YYYY-MM-DD | Date range |
python3 scripts/orchestrate-brief.py --detailed | Raise word cap ~150→~300/project |
python3 scripts/orchestrate-brief.py --persist-disk | Also write briefs to disk |
Range rollup: For --since or --range, emits ONE aggregated markdown per
project: single Executive Summary across the range, What Changed grouped by day,
Open Loops/Next Best Moves/Evidence aggregated.
Python Helpers (scripts/config.py)
The config.py module provides:
| Function | Returns | Description |
|---|
load_config() | envelope | Load (or bootstrap) config |
resolve_project(name_or_none) | envelope | Resolve one or all projects |
brief_path(project, date) | Path | Path for a brief file |
brief_exists(project, date) | bool | Whether brief exists on disk |
briefs_dir(project) | Path | Per-project briefs directory |
Multi-field functions emit
core/contracts/execution-result.schema.json
envelopes. Single atomic values (Path, bool) are returned bare.
Data Collection (scripts/query-sources.py)
The query-sources.py script aggregates raw data from three sources for a given
project and date:
python3 scripts/query-sources.py --project claude-code-plugins --date 2026-04-23
Output is a single JSON blob conforming to core/contracts/execution-result.schema.json
with these data fields:
| Field | Source | Description |
|---|
sessions | open-brain | session_summary + debrief entries |
closed_beads | beads (bd) | Beads closed in the date window, commits merged in |
open_beads | beads (bd) | Current open bead snapshot |
ready_beads | beads (bd) | Unblocked/ready bead snapshot |
blocked_beads | beads (bd) | Currently blocked beads |
commits | git | Standalone commits not linked to a bead |
learnings | open-brain | type=learning entries |
decisions | open-brain | type=decision entries |
decision_requests | beads (bd) | Beads flagged for human decision |
followups | open-brain | Lines with Decide: / Need input: / Follow-up: prefixes |
rework_signals | git + beads | Revert commits + supersede events |
warnings | — | Source unavailability notices (partial data) |
Status is ok when all sources are available, warning when any source is degraded.
CLI options:
--project NAME — project name from daily-brief.yml--date YYYY-MM-DD — date to query (Europe/Berlin midnight-to-midnight)--config PATH — optional config file path override
Capability Extraction (scripts/capability-extractor.py)
The capability-extractor.py script parses closed bead titles/descriptions and session
summaries for capability signals, producing short present-tense capability sentences
grounded in source data.
Capability signals detected:
- Bead titles/descriptions containing:
now, new gate, now verified, now possible,
unblocks, [FEAT], [QG] (case-insensitive)
- Both
feature AND task beads qualify
- Session summary lines starting with
New:, Fixed:, or Internal:
- For polaris: files created on target date in
docs/ and docs/adr/
CLI options:
--stdin — read query-sources.py JSON envelope from stdin--project NAME --date YYYY-MM-DD — run query-sources internally--scan-docs — enable polaris docs/ scanning--config PATH — optional config file override
Output: execution-result envelope with data.capabilities[] list of sentences.
Usage:
python3 scripts/capability-extractor.py --project claude-code-plugins --date 2026-04-23
Rendering (scripts/render-brief.py)
The render-brief.py script renders a v1.1 Chief-of-Staff markdown report in Voice B
(journalistic narrator, past tense, third-person observational, German).
v1.1 sections rendered:
| Section | German heading | Content type |
|---|
| Executive Summary | ## Executive Summary | Synthesized prose (German, ~150 words) |
| What Changed | ## Was sich verändert hat | Deterministic from closed_beads + commits |
| Why It Matters | ## Warum es zählt | Synthesized prose, only from cited facts |
| Open Loops | ## Offene Fäden | Deterministic from open_beads + blocked_beads |
| Decisions Needed | ## Entscheidungsbedarf | v1.1 Explicit sources only — see below |
| Drift & Rework | ## Drift- und Rework-Signale | v1.1 Deterministic from rework_signals |
| Next Best Moves | ## Nächste sinnvolle Schritte | Max 3 items: ≤2 from ready_beads, ≤1 from followups |
| Evidence | ## Belege | Bullet list: beads/commits/sessions/decisions/warnings |
Voice B rules: German, past tense, third-person observational. No first-person pronouns.
Prose paragraphs, not bullet lists (except Evidence). Empty day: Ruhiger Tag — keine Aktivität verzeichnet.
Entscheidungsbedarf (Decisions Needed) — v1.1
Renders only when explicit source signals are present. NEVER infers decisions from commits,
closed beads, session prose, or any other implicit source.
Explicit sources:
decision_requests — beads returned by bd human list --json (flagged for human decision)decisions — open-brain entries where entry["metadata"]["status"] == "pending"followups — entries with type == "Decide" or type == "Need input" ONLY
(entries with type == "Follow-up" are excluded — they belong in Next Best Moves)
Source tags: Each item cites *(Quelle: decision-request)*, *(Quelle: ob-decision)*,
or *(Quelle: followup-decide)*.
Empty state: "Keine offenen Entscheidungen erfasst."
Drift- und Rework-Signale (Drift & Rework) — v1.1
Renders only when rework_signals are present. Sources are deterministic from the
rework_signals field of the query-sources.py envelope — no other sources are used.
Signal types:
revert_commit — a git revert commit was detected; renders as \sha` subject (Quelle: revert-commit)`supersede_event — a bead was superseded; renders as **bead_id**: title *(Quelle: supersede-event)*reopen_event — a bead was closed and then reopened; renders as **bead_id**: title *(Quelle: reopen-event)*
Semantic distinction from Open Loops:
## Offene Fäden = beads currently in-flight (open or blocked but not finished)## Drift- und Rework-Signale = work that REGRESSED (finished work reversed, or scope changed)- These are mutually exclusive: a bead from Open Loops never appears in Drift, and vice versa.
Empty state: "Keine Drift- oder Rework-Signale erkannt."
CLI options:
--project NAME — project name from daily-brief.yml (required)--date YYYY-MM-DD — single day mode--since YYYY-MM-DD [--until YYYY-MM-DD] — range mode--range START END — explicit date range--detailed — raises word cap to ~300 words/project; full per-day sections in range--no-persist — skip writing brief to ~/.claude/projects/<slug>/daily-briefs/YYYY-MM-DD.md--config PATH — optional config file override
Single day:
python3 scripts/render-brief.py --project claude-code-plugins --date 2026-04-23
Config CLI Usage
python3 core/skills/daily-brief/scripts/config.py load
python3 core/skills/daily-brief/scripts/config.py resolve mira
python3 core/skills/daily-brief/scripts/config.py brief-exists mira 2026-04-24
Reference Docs
Sample Output
