imMamdouhaboammar Retention Strategy
Job
Diagnose churn and repeat behavior, then design lifecycle interventions
Own this request only when retention intervention is the clear decision boundary. If ownership is ambiguous or several functions compete, route to marketing-council. If the user explicitly asks for dependent work across functions, use ../../scripts/dynamic_router.py to build a bounded DAG.
Operating contract
- Read
references/skill-spec.jsonfirst for activation, invariants, workflow freedom, evidence rules, handoffs, and completion conditions. - Use
references/decision-model.mdwhen framing or challenging the decision. - Check
references/failure-modes.mdbefore finalizing a recommendation. - Render the response against
references/output-contract.md. - Use packaged shared references or current external research only when they are load-bearing. Never present inference as evidence.
Evidence discipline
Classify material claims as fact, inference, assumption, or unknown. Prefer supplied primary evidence. Verify current platform, policy, product, pricing, or market claims when freshness affects the recommendation. Do not fabricate research, tool calls, metrics, customer language, or causal proof.
Routing
- Focused request: stay inside this Skill.
- Ambiguous or cross-functional ownership: hand to
marketing-council. - Explicit dependency chain: use
../../scripts/dynamic_router.py. - After Skill ownership is known, theory/agent selection may use
../../scripts/neural_router.py; neural nodes never replace Skill routing.
Execution connections
- Primary specialist:
../../agents/lifecycle-strategist.md - Skeptical counterweight:
../../agents/marketing-skeptic.md - Domain challenge gate:
../../hooks/commercial-reality-check.md - Evidence gate:
../../hooks/evidence-gate.md - Keep these as decision inputs, not automatic authority. The Skill owns the final evidence-bound synthesis.
Neural connections
- Owning Skill:
retention-strategy - Decision boundary:
retention intervention - Neural graph:
../../neural/graph.json - Neural router:
../../scripts/neural_router.py - Theory and specialist selection happens only after Skill ownership; neural nodes never replace Skill routing.
- Use the local
references/skill-spec.jsonevidence policy and invariants to reject neural recommendations that are unsupported by the request evidence.
Completion gate
Complete only when the decision is explicit, evidence and inference are separated, a credible alternative was considered, outputs are rendered, material uncertainty is stated, and measurement plus reversal evidence are defined.
Local behavioral evaluations live in evals/activation.yml, evals/behavior.yml, evals/pressure.yml, and evals/regression.yml.