master-performance-assessment

Use when running any client performance assessment — new client onboarding (standard mode), monthly retainer check (quick mode), or full cross-domain deep audit with live API pulls and competitor benchmarking (deep mode). Replaces client-audit and monthly-pulse. Also use when the user says “audit my site”, “monthly check”, “new client”, “retainer report”, “deep dive”, “full assessment”, or “assess performance”.

Model	Source
sonnet	pack: assessment

Full Reference

Master Performance Assessment

The unified assessment orchestrator. Three depth modes cover every scenario — from a 30-minute monthly pulse to a full 4-8 hour cross-domain deep dive with competitor benchmarking.

Replaces: client-audit (→ standard mode) and monthly-pulse (→ quick mode)

Quick Reference

Mode	Duration	When	Spokes
`quick`	~30 min	Monthly retainer, trend monitoring	4 parallel
`standard`	~2–4 hrs	New client onboarding, quarterly deep-dive	6–7 parallel
`deep`	~4–8 hrs	Full cross-domain audit, competitor benchmarking	8 parallel + synthesis

Item	Value
Output file	`reports/assessment-[mode]-[YYYY-MM-DD].html`
History file	`performance-history.json` at project root
Tracking gate	`standard` and `deep` modes: tracking must be PASS or PARTIAL before continuing
Approval gate	All modes: client approves findings before action plan is built
Local spoke	Conditional — local businesses only (physical address + GBP relevant)
Graceful degradation	Missing packs → skip their spoke with ⚠ warning, don’t hard-fail

Reference Index

I want to…	File
See which spokes run per mode, how to dispatch them, conditional logic	`reference/mode-configs.md`
Understand how spoke dispatch works (parallel subagents, data handoff)	`reference/spoke-dispatch.md`
Implement the cross-domain correlation + prioritization synthesis engine	`reference/synthesis-engine.md`
Generate the HTML report (sections, tabs, templates, health scores)	`reference/report-generation.md`

Usage: Read the reference file matching your current phase. Each file is self-contained.

Topology

Hybrid pipeline + star swarm. Three phases:

Phase 1: COLLECT  →  Star (parallel spokes — each invokes 1–4 skills)
Phase 2: SYNTHESIZE  →  Pipeline (sequential: correlate → prioritize → recommend)
Phase 3: DELIVER  →  Single agent (HTML report + exec summary + action plan)
Phase 4: CONTINUE?  →  Prompt: brainstorm all fixes in priority order?

Read reference/spoke-dispatch.md for Phase 1 mechanics. Read reference/synthesis-engine.md for Phase 2 logic. Read reference/report-generation.md for Phase 3 templates.

Mode Selection

If no mode is specified, ask ONE question:

▸ Quick check (~30 min) for trend monitoring, or standard audit (~2–4 hrs) for a deep dive?

Request Pattern	Mode
”monthly check”, “retainer report”, “quick pulse”, “how’s the site doing”	`quick`
”new client”, “onboard”, “audit my site”, “full audit”, “quarterly review”	`standard`
”deep dive”, “everything”, “all data”, “competitor benchmarking”, “full assessment”	`deep`

After Assessment

Save assessment to performance-history.json via performance-memory (snapshot.create), then prompt:

Full assessment complete. Want me to run a brainstorm on addressing every issue in priority order?

If yes → invoke brainstorming with full findings + prioritized issue list as input context.

Announcement

┏━ 🎯 master-performance-assessment ━━━━━━━━━━━━━━┓ ┃ [mode] assessment — cross-domain performance ┃ ┃ intelligence with trend context ┃ ┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛