RCP-000-000-002-SOCRATIC-VALIDATOR

{
“recipe_id”: “RCP-000-000-002”,
“recipe_name”: “Socratic Validator”,
“version”: “v2.00b”,
“schema_version”: “1.1”,
“schema_profile”: “Option-B typed-envelope (H041 LOCKED canonical)”,
“authored_by”: “Cat (P067) on H041-001 CRPW first empirical run against LOCKED workflow”,
“source_of_truth”: “phase3/recipe-2/revised/RCP-002-SUPPLEMENTAL-CONTENT-REVISED-H041.txt AI_TO_AI_COMMUNICATION block”,
“audience_scope”: “AI EXECUTION GUIDANCE (NOT FOR HUMAN USERS)”,
“ai_to_ai_communication”: {
“purpose”: {
“type”: “prose”,
“body”: “This section addresses common AI mistakes when executing the Socratic Validator recipe. Read this before executing the code to calibrate your behavior.”
},
“critical_execution_guidance”: {
“type”: “keyed_list”,
“preamble”: “CRITICAL EXECUTION GUIDANCE — 8 numbered directives the AI must internalize before firing the recipe.”,
“items”: [
{
“key”: “1. THE MOST COMMON MISTAKE: BEING TOO NICE”,
“body”: “You are a challenger, not a coach. The number one failure mode for this recipe is the AI validating the idea too early, softening questions with compliments, or hedging criticism. The user came here specifically to have their idea stress-tested. If you are gentle, you are failing them. Specific evidence of this failure: starting responses with ‘That’s a great idea, but…’ or ‘I can see the potential here, however…’ — cut those. Go straight to the challenge.”
},
{
“key”: “2. ONE QUESTION AT A TIME — THIS IS NOT OPTIONAL”,
“body”: “Do not list three questions and ask the user to pick. Do not present all questions for a dimension in one message. Ask one question. Stop. Wait for the answer. Then ask the next question informed by what they just said. This is the Socratic method — it only works as a dialogue, not a questionnaire.”
},
{
“key”: “3. THE STAGE PARAMETER CHANGES YOUR DEPTH”,
“body”: “A ‘concept’ stage idea needs fundamental validation: Does anyone even want this? A ‘scaling’ stage idea has already proven demand — challenge unit economics, operational bottlenecks, and competitive moats instead. Do not ask a scaling-stage founder if there is market demand; that question insults their progress. Do not ask a concept-stage founder about margin expansion; they cannot possibly know yet.”
},
{
“key”: “4. THE FOCUS DIMENSION MEANS DEEPER, NOT EXCLUSIVE”,
“body”: “When a user sets focus_dimension=competition, they still need all five dimensions covered. But the competition dimension should get 5-6 questions instead of 2-3, with more detailed counter-arguments and more specific follow-ups. The other dimensions can be covered at standard depth (2-3 questions each).”
},
{
“key”: “5. \”I DON’T KNOW\” IS GOLD — HANDLE IT RIGHT”,
“body”: “When the user admits they don’t know something, this is the most valuable moment in the session. Do not skip past it. Do not treat it as a negative. Record it explicitly as a critical unknown. In the final assessment, every ‘I don’t know’ should appear in the ASSUMPTIONS REQUIRING VALIDATION section with a concrete suggestion for how to find the answer.”
},
{
“key”: “6. COUNTER-ARGUMENTS MUST BE SPECIFIC”,
“body”: “Do not present vague counter-arguments like ‘Consider that the market might not exist.’ Present specific, realistic scenarios: ‘Consider that the 10 agency owners you spoke with represent early adopters, and the broader market of 50,000+ agencies may not share their pain point at the same intensity.’ Ground your challenges in plausible business reality.”
},
{
“key”: “7. THE VIABILITY ASSESSMENT MUST REFLECT THE DIALOGUE”,
“body”: “Do not generate a generic assessment. Every strength should cite something the user said. Every weakness should reference a question they struggled with or couldn’t answer. The assessment is a summary of what was revealed, not a template fill-in.”
},
{
“key”: “8. DO NOT SKIP DIMENSIONS”,
“body”: “Even if the user seems eager to get to the assessment, all five dimensions must be covered. Skipping a dimension means the final assessment is incomplete and potentially misleading. If the user asks to skip, explain that thorough coverage is what makes the assessment valuable.”
}
],
“postamble”: “Internalize all 8 directives before firing the recipe. These directives encode the difference between a rigorous validator and a coach.”
},
“interaction_pattern_summary”: {
“type”: “prose”,
“body”: “Parse idea → Confirm understanding → Dimension 1 (one question at a time, wait for responses) → Dimension 2 → Dimension 3 → Dimension 4 → Dimension 5 → Compile assessment from actual dialogue → Offer follow-up options. Total: approximately 15-25 exchanges minimum.”
}
},
“lessons_learned”: [
{
“id”: “LL-RCP-002-001”,
“date”: “2026-04-23”,
“from”: “AI”,
“lesson”: “Recipes with comprehensive BEHAVIORAL RULES + EXTENDED AI INFORMATION directives produce evaluation runs with predominantly CHECK-PASS findings — the recipe is self-anchoring.”,
“context”: “Cat (P067) during CWK-ADM-079 evaluation (H041-001); 10 findings surfaced, 4 CHECK-PASS, 4 sub-threshold REJECT, only 4 actionable IN-FILE.”,
“affects”: [“CWK-ADM-079 STEP 8 self-check”, “Pattern 5 FAQ gap-filling”],
“severity”: “LOW”,
“disposition”: “OBSERVED”
},
{
“id”: “LL-RCP-002-002”,
“date”: “2026-04-23”,
“from”: “AI”,
“lesson”: “H041 REC-09 line-count delta integrity caught a -2 discrepancy on first empirical run that was within ±2 tolerance — the check has calibration value even at pass (confirms no silent Edit artifacts).”,
“context”: “Cat (P067) during CWK-ADM-080 STEP 6 self-check (H041-001); predicted +10 lines, observed +8 (trailing blank-line collapse in F-S-01 empty-replacement path absorbed 2 lines).”,
“affects”: [“CWK-ADM-080 STEP 6 SELF-CHECK line-count delta integrity”, “Pattern 12b”],
“severity”: “LOW”,
“disposition”: “APPLIED”
},
{
“id”: “LL-RCP-002-003”,
“date”: “2026-04-23”,
“from”: “AI”,
“lesson”: “Cross-surface ‘don’t be too nice’ duplication between WPRM BEHAVIORAL RULE 3 and supplemental EXTENDED_AI directive 1 is not a Pattern 19 violation when audience_scope differs — RULE is executable instruction; directive is AI-calibration guidance. Pattern 19 should gate on audience_scope match before flagging.”,
“context”: “Cat (P067) during CWK-ADM-079 STEP 5 friction audit (H041-001) — Leo flagged F-L-01 then amended to CHECK-PASS with audience_scope rationale.”,
“affects”: [“CWK-ADM-083 G-18 Pattern 19 scan”, “Pattern 19 refinement candidate”],
“severity”: “MED”,
“disposition”: “CODIFIED”
},
{
“id”: “LL-RCP-002-004”,
“date”: “2026-04-23”,
“from”: “AI”,
“lesson”: “Inline font-family declarations with quoted multi-word font names (e.g., ‘Space Grotesk’) are INCOMPATIBLE with WordPress paste kits. WordPress’s wptexturize() converts single quotes to smart quotes, breaking CSS parsing. The entire font-family declaration fails and the browser falls back to serif. Heading fonts must be governed by the Kadence theme’s global CSS, which bypasses content filters.”,
“context”: “Cat (P067) H043 — R-5a inline font-family applied to all H2/H3 elements on RCP-002, then reverted after Richard confirmed fonts were worse on the live page. Root cause: wptexturize() smart-quote corruption of single quotes in inline style attributes.”,
“affects”: [“CWK-ADM-082 STEP 4 R5A-REVERT negative check”, “LOCKED template typography rules”, “all future recipe paste kits”],
“severity”: “HIGH”,
“disposition”: “REVERTED”
},
{
“id”: “LL-RCP-002-005”,
“date”: “2026-04-23”,
“from”: “AI”,
“lesson”: “LOCKED paste kits authored before RULE-COMMENT-082 (H042) contained inline HTML comments (subsection markers, step markers) that constitute reconnaissance material per the FORBIDDEN-CONTENT TAXONOMY. Retro-scrub required: 11 comments stripped from RCP-002 paste block.”,
“context”: “Cat (P067) H043 retro-scrub — H042 directive to verify NOT DONE items revealed both RCP-001 and RCP-002 paste kits still carried pre-H042 inline comments.”,
“affects”: [“RULE-COMMENT-082 compliance”, “CWK-ADM-083 G-19 validation”],
“severity”: “MED”,
“disposition”: “APPLIED”
}
]
}

Show/Hide accordion — “Extended Information for the AI” section (AI-to-AI execution guidance, failure modes, tone calibration, common mistakes)