Unassailable Output

Input: $ARGUMENTS

Overview

Every output must be unassailable. Anything said can and will be used against it. Therefore: nothing wrong with form, content, length, logic, precision, completeness, or necessity.

The test: If a critic examined this output looking for any flaw, would they find one? If yes, revise until no.

Steps

Step 1: Produce Initial Output

Write the first draft response to the input. Don’t over-think it — just get the content down.

Step 2: Attack on All Dimensions

Examine the output through each lens. For each, ask: “Could a critic find fault here?”

Form:

• Grammar, spelling, punctuation correct?
• Formatting consistent and appropriate?
• Structure logical (not random ordering)?
• Length appropriate (not padded, not truncated)?

Content:

• Every claim accurate?
• Every claim necessary (not filler)?
• Nothing important missing?
• No contradictions within the output?
• No ambiguity where precision is needed?

Logic:

• Every inference valid?
• No unstated assumptions doing work?
• No logical fallacies?
• Conclusions actually follow from premises?
• Alternatives genuinely considered (not straw-manned)?

Precision:

• Quantitative claims specific (not “many” when you mean “7”)?
• Qualifiers appropriate (not “always” when you mean “usually”)?
• Terms defined where they could be misread?
• Scope stated (not overgeneralizing)?

Completeness:

• All parts of the question addressed?
• Edge cases considered?
• Limitations acknowledged?
• Counter-arguments addressed?

Necessity:

• Every sentence earns its place?
• No redundancy?
• No performative hedging (“I think maybe perhaps…”)?
• No throat-clearing before the point?

Step 3: Catalog Vulnerabilities

For each flaw found:

Step 4: Fix All Vulnerabilities

Starting with critical, then major, then minor:

1. Apply each fix
2. Verify the fix doesn’t introduce new flaws
3. Re-check the dimension after fixing

Step 5: Adversarial Final Check

Adopt the perspective of:

1. A hostile critic — looking for any weakness to attack
2. A confused reader — looking for anything unclear
3. A fact-checker — looking for any inaccuracy
4. A logician — looking for any invalid inference
5. An editor — looking for any unnecessary word

If ANY of these find something → fix it and re-check.

Step 6: Output

Present the revised, unassailable output. Do not include the audit process unless explicitly requested — the output should simply BE correct, not explain how it became correct.

If the output cannot be made unassailable (e.g., insufficient information to verify claims), state what would be needed to make it so.

When to Use

• High-stakes outputs (published writing, critical decisions, formal reports)
• When the output will be scrutinized by adversaries
• Quality control on any important deliverable
• → INVOKE: /cv (convergent validation) for multi-method verification
• → INVOKE: /stc (steelmanned counterarguments) for adversarial testing

Verification

• All six dimensions checked (form, content, logic, precision, completeness, necessity)
• All found flaws cataloged with severity
• All critical and major flaws fixed
• Fixes verified not to introduce new flaws
• Adversarial final check passed
• Output is clean (audit trail removed unless requested)