Skill: decompose-priority

Strategic anchor → measurable outcomes. Owned by the Architect agent.

When to use

• A quarterly MIT exists but the path forward is fuzzy
• A weekly priority is too big for one execution block — needs to be sliced
• A new VFO was just confirmed and you want to lay out the workstreams under it
• After /synthesize-priorities, when ratifying the draft

Procedure

1. Read the priority verbatim from Context/Priorities.md. If priority is a wikilink, resolve it.

2. Load context appropriate to the priority:

   • For brand priorities → Brands/{brand}/BrandContext.md + voice-profile + revenue-model + compliance + knowledge-base
   • For product/project priorities → the project's _Project.md + linked SPEC.md
   • For role priorities → Brands/{brand}/Roles/{role}.md
   • For fitness/personal priorities → Departments/Personal/Fitness/_Fitness.md + relevant phase doc
   • Always: relevant Wiki/Libraries/*.md (use the library names referenced in the priority's wikilinks; query them via /library query if needed)

3. Generate count candidate Goals (default 3, max 5). Each candidate is:

   • Concrete — names a specific outcome ("RMBC2 Stage-1 research package complete for puravida"), not an aspiration ("understand the customer better")
   • Measurable — has an unambiguous "done" criterion
   • Time-boxed — fits within the priority's horizon (weekly ≤ 7 days, MIT ≤ 90 days, VFO ≤ 365 days, DSS ≤ 3 years)
   • Routed — has a hint at which agent / skill / template will execute it (the Plan comes later via /generate-plan)

4. For each candidate Goal, write Wiki/Goals/{kebab-slug}.md using the template:

---
type: goal
priority: <a class="wikilink wikilink-broken" href="#"><priority wikilink or quoted text></a>
priority_section: <dss | vfo | mit | weekly>
status: proposed                    # proposed → active → in-progress → done | abandoned
horizon: <weekly | 90-day | annual | 3-year>
created: YYYY-MM-DD
target_completion: YYYY-MM-DD
suggested_template: <rmbc2 | decision-memo | phased-build | freeform>
suggested_agents: [<agent-slug>, ...]
linked_artifacts:
  - <a class="wikilink wikilink-broken" href="#"><wikilink></a>
  - <a class="wikilink wikilink-broken" href="#"><wikilink></a>
brand: <slug-or-blank>
department: <slug-or-blank>
---

# Goal: <Display Name>

## Charter
(2–3 sentence statement of what's being achieved and why it advances the parent priority.)

## Definition of done
- (specific, measurable, observable outcome 1)
- (specific, measurable, observable outcome 2)
- (specific, measurable, observable outcome 3)

## Why this goal (rationale)
(1 paragraph: what was in the source priority, what context was loaded, why this Goal is the right slice.)

## Risks / open questions
- (item)
- (item)

## Next move
Run `/generate-plan` with this Goal selected. The Architect will produce a 1–5 step Plan with agent + skill assignments, which you can review/approve before execution.

5. Surface to Adam — return the list of Goal paths + a one-paragraph summary explaining the decomposition logic (why these N, why not other plausible ones).

6. Append an entry to Wiki/Log.md (agent: architect, skill: decompose-priority).

Hard rules

• Don't write a Goal without a clear "done" criterion. If you can't write one, it's research — return a "Research Goal" with the question to answer instead.
• Don't propose Goals that contradict Hard Avoids (compliance forbidden patterns, fitness sensitivities, family-first DSS rule).
• Never write directly to Brands/ or Departments/ from this skill. Goals live only in Wiki/Goals/. Other artifacts get created by Plans the Architect generates separately.
• Slug collision protection: if Wiki/Goals/{slug}.md already exists, append a date suffix -2026-05-04 rather than overwriting.

Cross-links

• Agents/Architect — owner
• Skills/synthesize-priorities — what produces the priorities this skill decomposes
• Skills/generate-plan — natural follow-up: turn each Goal into a reviewable Plan
• Wiki/Goals/ — where the output lands