Stakeholder Due

Compute and surface what's overdue for reflection. The output is a prioritized list of question × stakeholder pairs the user should consider tackling, plus quick links to start each.

Inputs

Default: scan everything. The user can narrow:

• "What's due in managing-up" → filter to one category.
• "What's due for John" → filter to one stakeholder.
• "What's due this week" → filter to items overdue by ≤ 7 days from their cadence.

Loading

1. Read ~/bettersense-work-reflections/stakeholders.json for the registered list.
2. Load questions.json from the stakeholder-reflect skill folder.
3. For each registered stakeholder, read their file. For each question whose stakeholder_categories matches the stakeholder's category, find the most recent dated entry under the matching ## <prompt> heading.

Computing overdueness

For each (stakeholder, question) pair:

• cadence_days = lookup of suggested_freq: weekly=7, biweekly=14, monthly=30, quarterly=90, biyearly=180.
• last_entry_date = most recent ### YYYY-MM-DD under that question's heading. If never answered, treat as never.
• days_since_last = today − last_entry_date (or ∞ if never).
• due_ratio = days_since_last / cadence_days. (≥1.0 means due; ≥1.5 means notably overdue; ≥2.0 means very overdue.)

Honor cadence_overrides in the stakeholder file's frontmatter if present (per-question or per-stakeholder cadence).

Prioritizing

Sort by due_ratio descending. Then apply two adjustments:

1. Cap items per stakeholder. Show at most 3 questions per stakeholder in the output — otherwise a single neglected stakeholder dominates and the user disengages.
2. Cap total items. Default 10. The user can ask for more, but starting with 10 keeps the list actionable.
3. Surface "never answered" items distinctly. These are not "overdue" in the usual sense — they're new commitments. Group them in a separate section so the user can decide if they want to take on a new question or stick to refreshing existing ones.

Output format

Stakeholder reflections due (as of <today>)

🟥 Very overdue (≥2× cadence)
  • [John Adams · managing-up] What are their priorities, incentives and constraints?
    last logged 2026-02-15 · biweekly cadence · 11 weeks since
  • ...

🟧 Overdue (≥1.5× cadence)
  • [Jill Smith · managing-across] What's not going well?
    last logged 2026-04-08 · biweekly cadence · 23 days since
  • ...

🟨 Due (≥1× cadence)
  • ...

🆕 Never answered
  • [Client Delivery ABC · teams] Are goals clearly defined?
    quarterly cadence · never logged
  • ...

To pick one: "let me reflect on [stakeholder]" or run `stakeholder-reflect`.

Use the colored markers consistently. They make the output skim-able at a glance.

When nothing is due

Say so plainly. Don't manufacture work.

"Nothing significant is overdue right now. Earliest item due is John's 'priorities, incentives and constraints' in 4 days. If you want to get ahead, that's the one."

Wiring to a schedule

When the user wants this to fire automatically, guide them based on what they're using:

Claude Code Desktop app (recommended): Go to Routines → New routine → Local. Set the schedule to Weekly, Monday, and use these instructions: Run /stakeholder-due and show me the list. Desktop Routines have full local file access, persist indefinitely, and support any cadence.

Terminal / CLI users: Use OS-level scheduling (cron on macOS/Linux, Task Scheduler on Windows). Session-scoped tasks created by asking Claude in the terminal expire after 7 days and are unsuitable for weekly cadences.

Never suggest cloud routines — they run on Anthropic's servers and cannot access ~/bettersense-work-reflections/.

Don't set up a schedule automatically — the user owns the cadence decision.

Operating principles

• The list is a menu, not an obligation. Frame the output as "here's what you could pick up." Pressuring the user to clear all due items is how the practice dies.
• Surface stakeholders the user has been avoiding. If a stakeholder has had zero entries in 4+ weeks despite multiple due items, call that out gently in the output. Avoidance is itself signal.
• Don't over-engineer the math. Simple days_since / cadence_days is enough. Resist the temptation to add weights, recency bonuses, or "importance scores" — those make the output feel less like a real assessment.
• Respect the user's "skip" or "not now" instinct. If they consistently skip the same question for the same stakeholder, that's worth surfacing as an open question, not silently re-listing.

Anti-patterns to flag

• Outputting the same flat list every time. If the user runs this daily, the list barely moves day-to-day. Suggest a less frequent cadence (weekly is the sweet spot).
• Treating teams the same as individuals. Teams often deserve a longer cadence (the situation moves slower). If cadence_overrides aren't set on a team, gently suggest setting one.
• Listing 30 items. A long list is a list nobody reads. Default to 10.
• Sorting purely by date. A weekly-cadence question 8 days late should rank above a quarterly-cadence question 30 days late. Always sort by due_ratio, not raw days.