hrd-hard-mode-analysis

name	hrd - Hard Mode Analysis
description	Apply maximum rigor to a problem — no shortcuts, no defaults, no simplifications. Every assumption surfaced, every edge case considered, every tradeoff explicit. For when getting it wrong is expensive.
output	{"format":"prose"}

Hard Mode Analysis

Input: $ARGUMENTS

---

Core Principles

1. Hard mode means no invisible assumptions. Every assumption must be surfaced, stated, and evaluated. If the analysis rests on "X is probably true," hard mode requires specifying what "probably" means (70%? 95%?), what changes if X is false, and how to verify X.

2. Hard mode means no shortcuts. Easy mode provides defaults. Hard mode makes you choose. Simple mode skips edge cases. Hard mode enumerates them. This isn't complexity for its own sake — it's for situations where the shortcuts could be the thing that causes failure.

3. Hard mode is justified by stakes. Use hard mode when the cost of being wrong is high: irreversible decisions, high-stakes commitments, safety-critical systems, situations where you'll live with the consequences for years. Don't use it for lunch choices.

4. Completeness is the goal, not difficulty. Hard mode isn't hard because it's confusing — it's hard because it's thorough. Every step should be clear. The difficulty comes from the number of things to consider, not from the opacity of the analysis.

5. Hard mode produces a decision audit trail. Every conclusion must be traceable to evidence and reasoning. Someone reviewing the analysis should be able to see exactly why each conclusion was reached, what was considered and rejected, and where uncertainty remains.

---

Phase 1: Scope and Stakes

[H1] PROBLEM: [the problem or decision]
[H2] STAKES: [what happens if you get this wrong — be specific]
[H3] IRREVERSIBILITY: [how hard is it to undo this decision?]
[H4] TIME_PRESSURE: [how much time for analysis? hard mode takes time]
[H5] HARD_MODE_JUSTIFIED: [yes/no — is the rigor warranted by the stakes?]

→ If hard mode isn't justified, say so and route to /ezy or /smpl.

---

Phase 2: Assumption Audit

Surface every assumption the analysis rests on.

[H-N] ASSUMPTION: [stated assumption]
     SOURCE: [where this assumption comes from — data, experience, convention, guess]
     CONFIDENCE: [0-100% with reasoning]
     IF_WRONG: [what changes if this assumption is false]
     VERIFICATION: [how to test this assumption]
     STATUS: [verified | plausible | uncertain | untestable]

Required Assumption Categories

Category	What to Check
Factual	Data points, statistics, reported facts
Causal	"X causes Y" — is the mechanism confirmed?
Scope	"This applies in our context" — does it?
Temporal	"This will still be true in [timeframe]"
Behavioral	"People/systems will respond by [doing X]"
Resource	"We have/can get [resource]"
Environmental	"The context won't change significantly"

---

Phase 3: Full Option Enumeration

List ALL options, not just the obvious ones.

[H-N] OPTION: [option name]
     DESCRIPTION: [what this option entails — specifically]
     REQUIREMENTS: [what must be true for this option to work]
     BEST_CASE: [outcome if everything goes well]
     WORST_CASE: [outcome if things go badly]
     EXPECTED_CASE: [most likely outcome with reasoning]
     EDGE_CASES: [unusual situations where this option behaves differently]
     SECOND_ORDER: [downstream effects beyond the immediate outcome]
     IRREVERSIBILITY: [how hard to undo if this option fails]
     DEPENDENCIES: [what this option depends on — people, resources, timing]

Option Completeness Check

• "Do nothing" considered as an option
• "Do the opposite of instinct" considered
• Options from analogous situations in other domains considered
• Combinations of options considered
• Partial versions of options considered

---

Phase 4: Risk Enumeration

For the most promising options, enumerate all risks:

[H-N] RISK: [risk description]
     PROBABILITY: [low | medium | high — with reasoning]
     IMPACT: [low | medium | high | catastrophic — with reasoning]
     DETECTION: [how would you know this risk is materializing?]
     MITIGATION: [what reduces probability or impact]
     CONTINGENCY: [what to do if it happens]
     RESIDUAL_RISK: [risk remaining after mitigation]

Risk Categories to Check

Category	Examples
Known risks	Identified and quantifiable
Known unknowns	Identified but not quantifiable
Failure modes	Specific ways each option could fail
Interaction risks	Risks from combining factors
Timing risks	Risks from doing this now vs. later
Reputation risks	How this looks to others
Precedent risks	What this commits you to in the future

---

Phase 5: Tradeoff Matrix

[H-N] TRADEOFF: [option A] vs [option B]
     A_ADVANTAGE: [where A is better]
     B_ADVANTAGE: [where B is better]
     DECISIVE_FACTOR: [what determines which advantage matters more]
     PREFERENCE_IF: [choose A if..., choose B if...]

For the top 2-3 options, construct:

TRADEOFF MATRIX
                    | Option A | Option B | Option C |
  Best case         |          |          |          |
  Worst case        |          |          |          |
  Expected case     |          |          |          |
  Irreversibility   |          |          |          |
  Time to result    |          |          |          |
  Resource cost     |          |          |          |
  Risk level        |          |          |          |
  Upside potential  |          |          |          |

---

Phase 6: Decision and Audit Trail

HARD MODE ANALYSIS
==================

PROBLEM: [stated]
STAKES: [why hard mode]

ASSUMPTIONS (verified):
  [list verified assumptions with confidence levels]

ASSUMPTIONS (uncertain):
  [list uncertain assumptions — these are the biggest risk]

OPTIONS CONSIDERED: [N total]
  ELIMINATED: [options and why each was eliminated]
  FINAL CANDIDATES: [2-3 with tradeoff summary]

RISKS:
  [top risks with probability × impact ranking]

RECOMMENDATION: [choice with full reasoning]
  BECAUSE: [specific reasons, traced to evidence]
  DESPITE: [acknowledged downsides]
  CONTINGENCY: [what to do if it goes wrong]
  REVIEW_TRIGGER: [what would cause you to revisit this decision]

CONFIDENCE: [level with specific sources of uncertainty]
AUDIT_TRAIL: [every major judgment call, what was considered, what was rejected]

READY FOR:
- /ezy — if this was too much and you want the simple version
- /soph — for structural analysis of the chosen option
- /ar — to explore what follows from the recommendation

---

Failure Modes

Failure	Signal	Fix
Analysis paralysis	Hard mode produced so much analysis that no decision was reached	The goal is a DECISION, not a document. Force a recommendation
Unjustified rigor	Hard mode applied to a low-stakes reversible decision	Check stakes first. Route to /ezy or /smpl if unwarranted
Assumptions unsurfaced	Analysis proceeded on hidden assumptions	Run the assumption audit categories. Every category must be checked
Options too narrow	Only considered 2-3 obvious options	Run the completeness check. Consider opposites, combinations, partials
Risk blind spots	Only considered technical risks, missed reputational or precedent risks	Check all risk categories
No contingency	Recommendation made but no plan for failure	Every recommendation needs: what if I'm wrong?
False precision	"73.2% chance of success" without basis	Confidence levels must be justified. Don't quantify beyond your actual knowledge

---

Depth Scaling

Depth	Assumptions	Options	Risks	Tradeoffs
1x	Top 5	3-4	Top 3	Key tradeoff
2x	Full audit	5-7	Full enumeration	Full matrix
4x	Full + sensitivity analysis	All + combinations	Full + interaction risks	Full + scenario analysis
8x	Full + adversarial challenge	Exhaustive	Full + pre-mortem	Full + decision tree

Default: 2x. These are floors.

---

Pre-Completion Checklist

• Stakes justified hard mode
• All assumption categories checked
• Each uncertain assumption has an if-wrong analysis
• "Do nothing" considered as an option
• Options include non-obvious alternatives
• All risk categories checked
• Tradeoffs between top options made explicit
• Recommendation includes contingency plan
• Review trigger specified (when to revisit)
• Audit trail is complete (every judgment call documented)

---

Integration

• Use from: high-stakes decisions, irreversible commitments, safety-critical analysis
• Routes to: /ar (explore consequences of recommendation), /soph (structural analysis)
• Complementary: /ezy (opposite — minimum cognitive load path)
• Differs from /ezy: ezy minimizes complexity; hrd maximizes thoroughness
• Differs from /soph: soph pursues structural insight; hrd pursues completeness and rigor
• Differs from /certainty: certainty pursues maximum confidence on truth claims; hrd pursues maximum rigor on decisions
• Differs from /dcp: dcp is a decision-making framework; hrd is a rigor level applied to any analysis