Live Jessie skill

90-Day Plan / Strategy Planning.

Uses approved context, active subscriptions, Harvest budget capacity, configured source assets, and consultant instructions to draft 90-day strategy and, after explicit approval, create implementation tasks in Asana.

When It Routes

Live Strategy skills/strategy-planning/SKILL.md

A task asks for a 90-day plan, strategy refresh, planning checkpoint, channel plan, or structured set of recommended next work.

Flow

Task intake

Jessie checks whether the request is strategy planning and whether the objective, client, and planning horizon are clear.

Toolkit

The resolved Business Unit scopes context retrieval, Harvest budget reads, Asana work context, configured read-only evidence tools, AI Gateway synthesis, and Asana write-back access.

Context

Jessie loads a bounded, source-diverse evidence capsule from approved client strategy, current priorities, constraints, active subscriptions, source assets, and task-specific consultant notes. Context pack IDs, manifest paths, and generated Agent Context indexes are provenance only, not reviewed evidence.

Evidence

Jessie uses Harvest project, budget-report, and effective budget-note evidence as the planning capacity source and may use configured channel evidence where the task calls for it.

Insight

The LLM synthesizes rationale and plan structure through Cloudflare AI Gateway.

Review output

Jessie posts a planning checkpoint or strategy artifact for consultant review.

Approved task creation

If the consultant explicitly approves the strategy and asks Jessie to create tasks, Jessie creates eligible Asana implementation tasks with stable work-key markers and reports created, reused, and skipped items.

Context Jessie Looks At

Approved client strategy, current priorities, client-requested changes, business constraints, channel rules, and source references.
Consultant notes explaining focus, budget, timeline, client appetite, known tradeoffs, or topics to avoid.
Linked parent, strategy, report, or prior planning task context.

Data And Tools

Asana task brief and related work records.
Approved client context and retrieved context chunks/facts.
Harvest project, budget-report, and project-note evidence for subscription capacity, dated budget reductions, backlog, carryover, transfers, and budget caveats.
Configured read-only evidence sources such as GA4, Search Console, Google Ads, or Meta when relevant and available.

LLM Insight Stage

Strategy planning requires AI Gateway synthesis. The LLM turns context and evidence into rationale, options, questions, and a draft plan. Task creation is deterministic post-approval Asana population, not an unreviewed model write.

Outputs

Strategy or 90-day planning artifact for consultant review.
Client-context evidence check showing whether client priorities, agency commitments, budget/capacity, subscription scope, performance context, and operational constraints were satisfied, missing, or unavailable.
Harvest-backed capacity summary and budget caveats for internal staff review.
Budget/backlog reconciliation showing normal monthly subscription budget, available backlog or carryover capacity over the planning period, proposed plan consumption, and estimated remaining capacity.
Clarifying questions when the objective, constraints, or approval gate is ambiguous.
Asana review comments lead with a readable summary, current position, core objectives, recommended focus, and planned work before supporting detail and trace references.
Optional proposed delivery tasks as a prompt for human approval, with explicit approval able to create eligible Asana tasks idempotently.
Created task titles are plain action titles; Asana scheduling fields and the task body carry timing.
Created task descriptions start with client and actual Harvest project code/name, then provide objective, why the work matters, instructions, done-when acceptance criteria, dependency handoff instructions where relevant, timing, approval boundary, minimal source context, and a compact one-line ops reference.
Monthly reporting reserve recommendations expand into one task per strategy-period month, scheduled after each covered month closes, and the writable Asana Estimated time field is populated from planned hours when available.
Strategy-refresh input tasks are scheduled before the day-60 refresh draft so the refresh has its evidence and backlog dependencies available.
Analysis, mapping, and recommendation tasks are followed by the next practical implementation, task-prep, or validation work so known inefficiencies are not left idle for a later month.

Guardrails

No client-facing publishing without consultant approval.
No delivery-task creation until explicit consultant approval asks Jessie to create the implementation tasks.
No blocked, deferred, upsell, or duplicate-risk recommendation is created as a production task.
No metadata-dump task descriptions; raw process, duplicate, capacity, and source identifiers must not dominate the human-facing brief.
No treating a Context Pack Version ID, source manifest, generated Agent Context index, or Box scan as reviewed client context unless approved facts or prompt-ready non-index chunks were actually retrieved.
No generic optimisation plan that silently ignores explicit client-requested campaign, creative, landing-page, CRM, or scope changes found in reviewed context.
No analysis-without-action gaps unless the delay is explicit, approved, and explained in the task body.
No broad quarterly reporting bucket when the work is recurring monthly reporting; split it into independently assignable monthly tasks.
No external system changes or strategy changes treated as final without human review.
No Strategic Settlements-specific plan signals or task seeds for other clients.
No Harvest transfer or moved-budget note is counted as active monthly subscription budget.

Example Prompt

Jessie, draft a 90-day plan for Plunkett Homes focused on improving organic lead quality while keeping paid media efficiency visible.