gap-analysis

Gap Analysis Protocol

EXTREMELY_IMPORTANT: This is a MANDATORY protocol, not a suggestion. Follow every step. Do not skip steps. Do not combine steps. Do not summarize. Work through each gate in order.

You have produced a design, plan, or recommendation. Before presenting it as final, you MUST run all 7 checks below. Do not skip any. Report each result explicitly.

If your harness supports subagents, dispatch these checks to a fresh-context reviewer using agents/reasoning-reviewer.md — a reviewer that has not seen your design's justifications catches what self-review misses. Required at the DEEP tier; recommended whenever the decision is hard to reverse.

CHECK 1 — Inversion

Ask: "How would I guarantee this fails?"

List at least 3 specific failure modes; keep listing while distinct ones come to mind, stop when candidates turn contrived. For each, verify your solution prevents it. If any failure mode is not prevented, flag it as an open risk.

CHECK 2 — Second-order consequences

Ask: "And then what?" three times.

• 1st order: [the immediate intended effect]
• 2nd order: [the effect of that effect — often unintended]
• 3rd order: [the cascading consequence — often counterintuitive]

If 2nd or 3rd order effects are negative, state whether they're acceptable or require design changes.

CHECK 3 — MECE coverage

Ask: "Have I covered the entire problem space with no gaps and no overlaps?"

List the dimensions you've addressed. Identify any dimension you haven't addressed. Gaps are where surprises live.

CHECK 4 — Map vs territory

Ask: "Where does my model diverge from reality?"

List at least 3 things you simplified, assumed, or ignored — more if you know of them. For each, state the risk if reality differs from your model.

CHECK 5 — Adversarial

Ask: "How would a hostile actor exploit this? How would a careless user break this?"

Identify the weakest point. State whether it's acceptable or needs hardening.

CHECK 6 — Simplicity

Ask: "Is there a simpler solution I'm overlooking?"