Tim's Week 3 picks
One winner per adoption phase, in phase order — explore, pilot, scale, embed — plus an honorable mention. Tap any row to jump to the card.
The Roadmap Prioritizer
Click your agent ideas into the tier × autonomy matrix; the prioritizer sequences them Tier 1 first and lays out the roadmap.
The Memory Architect
Pick what your agent must remember; each need maps to one of the four memory types with its store and loop wiring.
The MCP Security Auditor
Grade your MCP server against the 12-item security checklist; the riskiest unaddressed gaps first.
Build or Don't Build? — The 6-Gate Agent Test
Six build-gates to a BUILD / PILOT / DON'T BUILD verdict with the first failing gate and its fix.
Where Are You on the Loop-Engineering Curve?
Five questions place you on the loop-engineering progression, with your next step and verification loops to add.
Bagent or Real Agent? — The PROOF Scorecard
The full 25-point PROOF rubric as an interactive self-evaluation with a Bagent-vs-Real-Agent verdict.
Would Your Agent Survive a Prompt-Injection Attempt?
Tick the 6 defense layers, pick the riskiest tool's tier, walk the 12-item checklist; grade A–F with gaps ranked by exploitability.
Is Your Agent's Context Budget Leaking?
Five diagnostic questions grade your SKILL.md setup A–F, plus the interactive 8-step refactor checklist.
Your First AI Incident Post-Mortem, Done Right
A guided blameless post-mortem builder with a scored detection-gap diagnosis and a copyable draft.
Multi-Agent Orchestration Pattern Picker
A 4-question picker recommending one of six orchestration patterns, which to avoid, and how your setup will fail.
AI Agent Profitability Auditor
Enter three numbers, get real cost per outcome with a Keep / Rebuild / Kill verdict plus five hidden costs.
LLM-Customization Decider
Walk the fine-tune vs RAG vs prompt engineering decision tree, get the recommended approach and the mistake to avoid.
Will Your Agent Skill Survive? — Penn's Skill Audit
A pre-release skill audit — 9 questions across his AI-level ladder and three eval gates, ending in SHIP IT / SHIP CAREFULLY / DO NOT SHIP.
Red-Team Your Own Agent — the 9-Question Pre-Launch Audit
His 9-question pre-launch checklist as an interactive audit with an attack-category coverage map.
Which coding model actually holds up in real use?
Three questions to a coding-model pick from one tester's paid SanityHarness runs — eval rank vs real-use quality.
Scorecard
Your ratings, ranked by total. Saved on this device.
Top pick per phase
Full ranking
| Drop | Phase | Uniq | Biz | Total |
|---|