# Monday Brief

> A Claude skill from actowery/claude-skills.

- **Type:** Skill
- **Install:** `agentstack add skill-actowery-claude-skills-monday-brief`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [actowery](https://agentstack.voostack.com/s/actowery)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [actowery](https://github.com/actowery)
- **Source:** https://github.com/actowery/claude-skills/tree/main/skills/monday-brief

## Install

```sh
agentstack add skill-actowery-claude-skills-monday-brief
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Monday Brief

A forward-looking Monday morning briefing: synthesizes recent signal across all sources (manager DMs, calendar, action items, people context, prior weekly reports) into a "what's top of mind this week" document. Tells the user what to do for themselves and the team, what's at risk, and what they owe.

**Distinct from `how-can-i-help`** (which is retrospective — surfaces what's already fallen through cracks) and from the weekly-update skills (which are output artifacts for others to read). This is **input**: the user reads it, takes notes, walks into the week with a plan.

## Config & cache locations

- **User config**: `${XDG_CONFIG_HOME:-$HOME/.config}/monday-brief/user.json` (auto-bootstrapped from `1on1-prep` config — same team roster)
- **Output**: `/monday-briefs/monday-brief-.md`
- **Research cache**: `${XDG_CACHE_HOME:-$HOME/.cache}/monday-brief//`

## Blast-radius table

| Action | Scope | When |
|---|---|---|
| Write `${XDG_CONFIG_HOME:-$HOME/.config}/monday-brief/user.json` | Local | Phase 0 init only |
| Write `/monday-briefs/monday-brief-.md` | Local | Phase 5 — path announced first |
| Write research cache locally | Local | Phase 3 |
| Read Outlook (email + calendar), Slack, Jira, GitHub, Confluence (action items page), local people-context files, prior weekly reports | Remote reads only | Phase 3 |
| Write to Slack, send email, comment on tickets, modify any remote resource | **Never** | — |

## Prerequisites

- Atlassian MCP (Jira + Confluence — for action items page)
- Slack MCP (DM-aware — `slack_search_public_and_private` mode)
- Outlook / M365 MCP (email + calendar)
- `gh` CLI authenticated with `repo` scope
- Existing `1on1-prep` / `action-item-sync` configs (for team roster + action items page ID)
- People context files at `/.md` (loaded per direct report)

If any source is missing, the skill still runs and notes the gap inline.

## Invocation

```
monday-brief                       # default — synthesize for the current/upcoming week
monday-brief           # explicit Monday date (e.g. 2026-05-25)
monday-brief --dry-run             # print without saving file
```

Also triggered automatically by a scheduled task (set up at Phase 6 — see "Scheduling" below).

## Workflow

### Phase 0 — Init / bootstrap config (only on first run)

**Side effects:** writes `${XDG_CONFIG_HOME:-$HOME/.config}/monday-brief/user.json`.

1. Check for existing config. If absent:
   - Bootstrap from `${XDG_CONFIG_HOME:-$HOME/.config}/1on1-prep/user.json` (display_name, atlassian_account_id, slack_user_id, github_username, team.members[], manager, peers[], pronouns).
   - Bootstrap action items page ID from `${XDG_CONFIG_HOME:-$HOME/.config}/action-item-sync/user.json`.
   - Pull workspace `output_dir` and `people_dir` from `${XDG_CONFIG_HOME:-$HOME/.config}/claude-skills/workspace.json` (see `_shared/workspace-config.md`).
2. Confirm or collect: `slack_search_mode` (default `public_and_private`), `prior_reports_lookback_days` (default 7), `output_dir` for briefs (default `/monday-briefs`).
3. Announce the XDG path and write the config.

### Phase 1 — Determine the week

**Side effects:** none.

1. Resolve the Monday in scope:
   - Bare invocation → today if today is Monday; otherwise the nearest upcoming Monday.
   - Explicit date → that Monday.
2. Window for synthesis = (previous Friday EOD) → (this Friday EOD) — i.e. the upcoming work week.
3. Lookback for signal = (this Monday minus `prior_reports_lookback_days`) → (this Monday) — captures recent DMs, action items, transcripts.
4. Announce: "Building Monday brief for week of ``. Signal window: `` → ``. Forward window: `` → ``."

### Phase 2 — Map this week's calendar

**Side effects:** read-only Outlook calendar call.

Pull the user's calendar for the week (Mon → Fri inclusive). Surface:
- **All scheduled 1:1s** with team members, manager, peers (extracted from `attendee_count == 2` and matched against `team.members[]` / `manager` / `peers[]`).
- **Big visible meetings** — anything with > 8 attendees, or with `presentation` / `review` / `demo` / `standup` keywords in the topic.
- **Anything new to this week** that wasn't on the calendar last Monday (only if a prior brief exists at `/monday-brief-.md`).
- **Time blocks already booked** — total hours of meetings this week, biggest free blocks.

### Phase 3 — Research signal sources

**Side effects:** read-only across all sources. Writes cache under `${XDG_CACHE_HOME:-$HOME/.cache}/monday-brief//`.

Announce sources up front:
> Building brief for ``. Sources: Outlook (calendar + email this week + recent from manager + recent unread), Slack (DMs from manager + DMs with team members + AI champion channels + recent mentions), Jira (open tickets owned by user + open tickets owned by team), GitHub (open PRs by team), Confluence action items page, people context files, prior weekly reports.

Run sources in parallel; cache raw results.

**Manager comms — highest signal.** Look for:
- Recent DMs from manager (last 7 days) — anything you owe a response to
- Recent group DMs that include manager
- Outlook emails from manager (last 7 days) — any direct asks
- **Heads-up signals** — explicit announcements of upcoming reviews, presentations, calibration cycles, leadership meetings. Phrases to watch for: "next week", "heads up", "FYI", "headed your way", "you'll need to", "by EOW", "ahead of", "before [date]".

**Action items page.** Parse `/tmp/action-item-sync-.adf.json` if present (recent run), otherwise fetch the page fresh. Filter to:
- Items where `owner == user.display_name` AND `status == Open` → user's pending commitments
- Items added in the last 7 days regardless of owner → recent commitments worth re-confirming
- Items still Open and older than 14 days → stale items at risk of slipping

**People context files.** Load every `/.md` for direct reports. For each person, extract:
- Career & Promotion (active goals — these drive what to surface in 1:1s this week)
- Development Focus (what to ask about / observe)
- Recent Notes (anything dated within the last 14 days that needs follow-up)
- Sensitive Context (informs tone, NEVER quoted)

**Jira.** Open tickets across the team. Particularly:
- Tickets owned by user (often planning / management commitments)
- Tickets owned by team members that haven't moved in 7+ days (potential blockers)
- Anything labeled `escalation` / `customer` / `urgent` / similar

**GitHub.** Open PRs by team members in configured orgs. Particularly:
- PRs open > 5 days (potential review-bottleneck signal)
- Draft PRs from > 3 days ago (planning vs stalled signal)

**Slack — beyond manager.** Search the AI champion channels (`#ai-champions`, `#team-devx-humans`, similar) for the last 7 days for:
- Anything tagged or mentioning user
- Cross-team announcements (skip-level / leadership posts)

**Prior weekly reports.** Read the latest:
- AI weekly update Confluence row (current user's row from last week — what was published)
- Weekly Confluence Platform Automation row (DevX block)
- Exec email (`/tmp/ai-exec-report-.eml` or recent in Outlook sent)
- These provide continuity — "what did I say last week was in flight that should now be done?"

**Optional: prior Monday brief.** If `/monday-brief-.md` exists, load it. Compare:
- What was top-of-mind last week — still relevant? Closed? Drifted?
- What risks did last week's brief name — did they materialize?

### Phase 4 — Synthesize the brief

**Side effects:** none — pure reasoning.

Produce a markdown brief following the structure in `references/output-style.md`. Summary:

**1. Top of mind for the week** — 2-4 things. Highest-signal items from manager DMs, calendar, action items. Be ruthlessly selective — these are the things the user *must* think about Monday morning. Cite the source explicitly (e.g. "from  DM ").

**2. Per-person — what to do for each report this week.** Table format. For each direct report, one row with:
- Their week ahead (what's on their plate, from Jira/GitHub/calendar)
- Your move (per-person action drawing on people context + recent signal)

Exclude direct reports who are out of office this week. Mark self-reporting / promotion-track people with a flag if there's promotion-relevant evidence to capture.

**3. Open action items you owe.** From the action items page where `owner == user` AND `status == Open`. Order by age (oldest first). For each, note what blocks closing it.

**4. Open items you owe responses on.** From Slack DMs / email threads from manager and key stakeholders where user owes a reply. Be careful to distinguish "owes a response" from "thread is active." Look for direct asks unanswered.

**5. Risks / things that could go sideways.** Forward-looking. What in flight could break this week? Anything from Phase 3 sources that suggests a risk pattern:
- A meeting outcome that hasn't been reported
- A team member showing fatigue / over-commitment
- A dependency that hasn't moved
- A scheduled review or presentation needing prep

**6. Quieter priorities to keep on radar.** Things that are sustaining but worth not forgetting. One-liner each.

**7. One-liner for the week.** A single sentence framing the dominant move of the week. This is the most-read line of the brief.

**Honest count rule:** if a section has no real content, omit it. Don't pad. Better to have 3 strong sections than 7 weak ones.

**Use the inline tag `(needs decision)`, `(overdue)`, `(at risk)`, `(in flight)` etc. liberally** — these scan well.

**People context safety:** Never quote `Sensitive Context` content. Use it only to adjust framing.

### Phase 5 — Deliver

**Side effects:** writes `/monday-brief-.md` and prints inline.

1. Save the brief to `/monday-brief-.md`.
2. Print the full brief inline as the primary delivery — this is what the user reads at 9 AM Monday.
3. End with a clickable sidebar link: copy to `/tmp/monday-brief-.md` and render as `[/tmp/monday-brief-.md](/tmp/monday-brief-.md)`.

After delivering: append a missing-sources note per `_shared/missing-sources.md` if any source was skipped or errored.

### Phase 6 — Scheduling

Triggered automatically each Monday via a scheduled task. To set up:

```
schedule create monday-brief --cron "30 8 * * 1" --tz "America/New_York"
```

Or via the `scheduled-tasks` MCP. Recommended timing: **Monday 8:30 AM in user's timezone**, so the brief is ready when the user sits down at 9 AM with coffee.

The scheduled run is identical to the manual invocation — same output path, same prints. The point is the brief exists when the user opens their laptop on Monday, not that it gets pushed to them.

## Safety rules

- **Read-only remote.** This skill never writes to Slack / email / Jira / GitHub / Confluence / calendar.
- **No fabricated emotion.** When framing per-person actions, refer to observable behavior, not invented internal state. The people-context files contain `Sensitive Context` — use that to adjust tone, never to quote.
- **Source citations required.** Every "top of mind" item must cite where the signal came from (Slack DM date, email subject, ticket key, calendar event). Without that, the user can't verify.
- **Honest count.** Empty sections get omitted. A short brief is a good brief.
- **Pronouns.** Per the people context file. If absent, use the person's name only. Never guess from name. See `_shared/pronoun-handling.md`.
- **Don't double up.** If a Monday morning calendar item already has a 1:1 prep brief at `/1on1-prep--.md`, the Monday brief references it but doesn't reproduce it.

## Files in this skill

- `references/data-sources.md` — query patterns for each signal source
- `references/output-style.md` — section structure, tone, citation format
- `config/user.example.json` — config template (real `user.json` is gitignored)

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [actowery](https://github.com/actowery)
- **Source:** [actowery/claude-skills](https://github.com/actowery/claude-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-actowery-claude-skills-monday-brief
- Seller: https://agentstack.voostack.com/s/actowery
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.