49 lines
3.2 KiB
Markdown
49 lines
3.2 KiB
Markdown
---
|
||
name: monitor-personalization
|
||
description: Audits personalization performance, governance compliance, and experiment results.
|
||
usage: /personalization-engine:monitor-personalization --initiative "PLG Onboarding" --window 14d --detail full
|
||
---
|
||
|
||
# Command: monitor-personalization
|
||
|
||
## Inputs
|
||
- **initiative** – personalization program or campaign to analyze.
|
||
- **window** – time frame (7d, 14d, 30d) for pulling metrics.
|
||
- **detail** – summary | full to control report depth.
|
||
- **dimension** – optional breakdown (profile, channel, cohort).
|
||
- **alert_threshold** – optional KPI threshold to trigger incident items.
|
||
|
||
### GTM Agents Pattern & Plan Checklist
|
||
> Mirrors GTM Agents orchestrator blueprint @puerto/plugins/orchestrator/README.md#112-325.
|
||
|
||
- **Pattern selection**: Monitoring usually runs **pipeline** (data aggregation → governance scan → experiment readout → issue detection → action plan). If governance + experiments review can run concurrently, capture a **diamond** block with merge gate in the plan header.
|
||
- **Plan schema**: Save `.claude/plans/plan-<timestamp>.json` capturing initiative, data feeds, dependency graph (data eng, privacy, experimentation), error handling, and success metrics (lift %, incident response time, consent adherence).
|
||
- **Tool hooks**: Reference `docs/gtm-essentials.md` stack—Serena for schema diffs, Context7 for governance/experiment SOPs, Sequential Thinking for retro cadence, Playwright for experience QA evidence.
|
||
- **Guardrails**: Default retry limit = 2 for failed data pulls or anomaly jobs; escalation ladder = Testing Lead → Personalization Architect → Data Privacy Lead.
|
||
- **Review**: Run `docs/usage-guide.md#orchestration-best-practices-puerto-parity` before distribution to ensure dependencies + approvals are logged.
|
||
|
||
## Workflow
|
||
1. **Data Aggregation** – pull engagement, conversion, and revenue impact by profile/channel plus decision tree health signals.
|
||
2. **Governance Scan** – verify consent flags, fallback rates, and rule change logs for compliance.
|
||
3. **Experiment Readout** – summarize live/completed tests with statistical confidence and recommended actions.
|
||
4. **Issue Detection** – flag anomalies (data freshness, variant suppression, performance dips) and suggest playbooks.
|
||
5. **Report Distribution** – publish recap with dashboards, backlog items, and owners.
|
||
|
||
## Outputs
|
||
- Performance dashboard snapshot segmented by profile/channel/variant.
|
||
- Governance checklist status with any violations or pending approvals.
|
||
- Experiment memo with next steps + rollout guidance.
|
||
- Plan JSON entry stored/updated in `.claude/plans` for audit trail.
|
||
|
||
## Agent/Skill Invocations
|
||
- `testing-lead` – interprets experiments and recommends rollouts.
|
||
- `personalization-architect` – validates experience integrity.
|
||
- `governance` skill – enforces policy checks and approvals.
|
||
|
||
## GTM Agents Safeguards
|
||
- **Fallback agents**: document substitutes (e.g., Governance covering Testing Lead) when leads unavailable.
|
||
- **Escalation triggers**: escalate if alert_threshold breached twice, consent violations appear, or anomaly alerts repeat; log remediation steps in plan JSON.
|
||
- **Plan maintenance**: update plan JSON/change log when metrics, thresholds, or monitoring cadences change to keep audits accurate.
|
||
|
||
---
|