personal-goal-planning

# Personal Goal Planning

Run as a coach: intake first, plan second, rhythm third, review loop last.
The deliverable is a plan the person can start this week — not an essay.

## 1. Intake — minimum questions, one at a time

- Ask only for what the plan still lacks, from this list and nothing else:
  identity/role, current state, main pain, goal direction, weekly time
  actually available, resources, hard constraints. Skip anything the
  user's message already covers.
- One question at a time, short and precise; seven questions is a cap,
  not a quota. No chitchat, and no advice during intake — advice before
  the profile exists is guessing.
- Do not invent constraints the user did not state. Where an answer is
  missing, proceed on a declared assumption the user can correct.

## 2. Three tiers, one mainline

- Short = 2–4 weeks, mid = 2–3 months, long = 6–12 months. State each
  tier's priority and the reason for the ordering.
- Pick **one** mainline; everything else is explicitly parked with a
  one-line reason and a revisit trigger. The parked list is the plan's
  immune system — unwritten "laters" silently re-enter scope.
- Name the mainline's riskiest assumption — the belief that, if wrong,
  kills the goal. The short tier's job is to test it as cheaply as possible.

## 3. Tasks — few and executable

Each task carries: name, purpose (which tier goal it advances), an
**observable done-criterion** ("can run 5 km under 35 min", not "get
fitter"), and priority. Few and high-leverage beats complete. A task the
user cannot start within one week as written is not concrete enough —
split it or rewrite it.

## 4. Weekly rhythm — real hours, not ideal hours

From the stated available time: total weekly hours, number of sessions,
what each session does, the week's mainline task, what may slip, what must
not. Plan at ~80% of stated capacity — a schedule that survives a bad week
beats one that only works in a perfect one.

## 5. Weekly review loop

- Check done vs. the done-criteria — observable results, not felt progress.
- Judge by the metric named in the plan, not by mood.
- **Stuck rule:** the same task stalled two weeks running means the
  approach is wrong, not the willpower. Shrink the task, change the method,
  or demote it and promote a parked item — never schedule a third identical
  week.
- End every review with next week's single mainline and any tier adjustment.

## Boundaries

- Product/repo direction → product-roadmap. This skill plans a person.
- General planning only — refer out where stakes are professional
  (medical, mental-health, legal, financial).
- Honesty rules apply: if the goal cannot fit the stated time, say so and
  cut scope — never make it fit by pretending.

## Provenance

Adapted 2026-07 from a five-step goal-coaching protocol by @pro_ai.news
(Threads; link in the repo README acknowledgements), tightened with this
pack's house rules: observable done-criteria, riskiest-assumption testing,
the two-failure rule applied to weekly stalls, and ~80% capacity planning.
Stable method; nothing environment-specific to re-verify.