adhd-daily-planner

This skill produces a morning brief: what changed overnight in the user's projects, what matters today, and one concrete next-action per project. On Monday it is the full weekly stack. On other weekdays it is a lighter daily-pulse.

Activation criteria

Activate when any of the following is true:

• The user types /resume monday or /morning-brief.
• The user's message contains one of: what's on today, plan my day, what should I do today, monday morning.
• This is the user's first interaction of a new local day AND preferences.output_format is answer_first (treat as an implicit invitation, not an obligation — see "Do not" below).

Do not activate inside a session that has already been running for more than thirty minutes; that conversation has its own context already.

Operating instructions

Follow these steps in order. Do not skip steps. Do not improvise additional tool calls.

1. Anchor the day. Call mcp-chronometric.get_time_context(). Note day_of_week, energy_zone, and current_session_length. If day_of_week is Monday, this is a weekly brief. Otherwise it is a daily-light brief.

1. Pull the weekly rollup, scoped or unscoped.

• For a weekly brief: call mcp-cognitive-graph.weekly_rollup() with no project filter to discover which projects had activity in the trailing seven days. Rank projects by the most recent decisions[].decided_on (falling back to most recent blockers[].recorded_at when there are no decisions).
• For a daily-light brief: same call, same ranking.
• Cap the project list at preferences.max_chunk_size (default 5). If more projects had activity, note the count of elided projects in the closing line — do not name them.

1. For each retained project, get its rollup. Call mcp-cognitive-graph.weekly_rollup(project=<name>) once per project. Use the returned decisions, blockers, and next_actions fields verbatim. Do not paraphrase decision names.

1. Decide whether to surface decision detail. If a project has decisions in the last seven days but weekly_rollup.next_actions is empty for that project, call mcp-cognitive-graph.recall_decisions(project=<name>, since=<thirty days ago>) once to surface up to two recent decisions that may need follow-up. Skip this call when next_actions is already populated.

1. Get one concrete next-action per project. Call mcp-task-fractionator.next_one(project=<name>) for each project. Three outcomes:

• Success with confidence >= 0.7 — quote the task title and the confidence value.
• Success with confidence < 0.7 — quote the title and explicitly say "low confidence" with the value. Do not pretend certainty.
• Error NO_TASKS_AVAILABLE — say "no decomposed tasks for this project; consider decompose if it's time to plan it." Do not fabricate a task.
• Any other error — note the error code in the project section and move on.

1. Render the brief. See "Output format" below.

1. Stop. Do not propose a calendar rearrangement, a follow-up question, or "shall I start the first task?". The brief is the deliverable.

Output format

Strict "Answer First" structure. The first sentence must fit in 80 characters and must name the day's shape in plain prose.

<One paragraph, ≤ 80 chars in the first sentence, plain prose. State which day it is,
the brief type (weekly or daily-light), and the count of projects covered. No
exclamation marks. No "let's crush it" register.>

### <Project name>
- Most recent decision: <decision name> (<decided_on>, conf <confidence>)
- Blocker: <blocker literal or "none">
- Next: <next_one.task.title> (<estimated_minutes> min, conf <confidence>)

### <Project name>
- ...

---
Elided <N> further projects with activity this week.
Energy zone right now: <energy_zone>. End-of-day stated as <chronometric.end_of_day_local>.
This brief is not a productivity scorecard. Yesterday's incomplete items are not graded.

Rules:

• Maximum three bullets per project. If a project has no blocker, write Blocker: none — do not invent one.
• Project sections appear in the order produced by step 2's ranking.
• Confidence is always rendered as a number with two decimal places, never as a word.
• The closing block is mandatory. The "not a scorecard" line grounds the LLM and the user.

If weekly_rollup() returned zero projects in the trailing seven days, emit the empty-graph response instead:

No projects in the last 30 days. Nothing to brief against.

If you'd like to start one, the next step is `mark_session_start(intent=<your intent>)`
in mcp-chronometric — that anchors today and gives tomorrow's brief something to draw on.

Distress signal handling

If the user's invoking message contains overwhelm phrases — I can't, too much, everything is on fire, I'm stuck, exhausted, burned out — reduce the project cap to 3 instead of preferences.max_chunk_size, omit the confidence numbers (state "high" / "medium" / "low" instead), and append one sentence to the closing block: If three is still too many right now, ask for "just one project" and I'll narrow further. Do not lecture. Do not diagnose.

Do not

• Do not use the words "superpower", "crusher", "smash", "let's go", "you got this", "differently abled".
• Do not enumerate yesterday's incomplete items.
• Do not compute or display a productivity score.
• Do not propose calendar buffer transitions unless the user's MCP environment surfaces calendar data through a tool you can call. Calendar integration is not in scope for v0.1.0 of this skill — if no calendar tool is wired, omit the "buffer transitions" feature silently.
• Do not call next_one for a project with no decisions and no blockers — that project does not need a next-action surfaced, it needs the user to decide whether to retire it.
• Do not invent project names. Only use names that came back from weekly_rollup.
• Do not exceed three bullets per project even if more data exists.
• Do not activate inside an ongoing >30 min session.

What this skill is not

It is not a productivity maximiser. It is not a guilt trip about yesterday. It is a once-a-day surface area for projects, decisions, and one tangible next thing.

Examples

See tests/01-monday-morning-brief.md, tests/02-empty-graph-fallback.md, and tests/03-multi-project-stack.md for the full invocation traces.