Yaksha: Скилл комплексного анализа и решения проблем

--- name: yaksha description: This skill should be used when the user asks to "analyze from multiple angles", "give me the full picture", "yaksha protocol", invokes "/yaksha", or faces a problem important enough to warrant multi-darshana analysis — running 2-3 philosophical reasoning engines on the same problem, then synthesizing through Vedanta and extracting actions through Mimamsa. --- # Yaksha Protocol — Multi-Darshana Convergence You are executing the Yaksha Protocol — named after the Yaksha's questions in the Mahabharata that could only be answered by holding multiple truths simultaneously. You will run 2-3 darshana engines on the SAME problem, find where they agree and where they pull apart, then synthesize and extract action. ## Procedure ### STEP 1: SELECT THE DARSHANAS Read the query. Select 2-3 darshanas (never more than 3) that will produce the most DIVERGENT analyses. The goal is not agreement — it is productive tension. Choose darshanas whose methods will see DIFFERENT things in the same problem. State your selection: ``` DARSHANA SELECTION: 1. [Name] — selected because it will reveal [what] 2. [Name] — selected because it will reveal [what] 3. [Name, if applicable] — selected because it will reveal [what] DROPPED: [any that were considered but rejected, and why] ``` ### STEP 2: RUN EACH DARSHANA FULLY For each selected darshana, execute its COMPLETE method. Do not abbreviate. Each darshana must produce at least ONE NON-OBVIOUS INSIGHT — something that would surprise a domain expert. If your insight is "this is a complex problem with multiple dimensions," you have produced nothing. An insight must be SPECIFIC, TESTABLE, and NOT the first thing someone would think of. Mark each insight clearly: **[NON-OBVIOUS INSIGHT]** Follow the full procedure for each: - **Nyaya**: 5-member syllogism, all 5 hetvabhasa checks, exact failure point, what it reveals. - **Samkhya**: 15+ components, causal chains, non-obvious components, completeness audit. - **Yoga**: All factors scored 0-10, ruthless cutoff at 7, dharana point (not the obvious answer), second-order focus, vritti audit. - **Vedanta**: Precise contradiction, adhyaropa (both fully true), apavada (negate projections), what remains, mahavakya reframing. - **Mimamsa**: Sentence classification, vidhi decomposition (agent/action/object/condition), dependency graph, execution sequence. - **Vaisheshika**: Recursive decomposition to atoms, 6-padartha classification, unexpected atom, essential vs accidental connections. ### STEP 3: IDENTIFY CONSENSUS AND TENSIONS After running all selected darshanas, map the results: **CONSENSUS:** Where do the darshanas AGREE? What conclusion do they all converge on, despite using completely different methods? Convergence from different methods is strong evidence. **TENSIONS:** Where do they DISAGREE or pull in different directions? State each tension precisely: - "Darshana A says [X], but Darshana B says [Y]. These conflict because [Z]." **HIDDEN SHARED ASSUMPTION:** What assumption do ALL the darshana analyses share that might be wrong? Every set of perspectives still shares a common frame — find it and name it. ### STEP 4: VEDANTA SYNTHESIS — The Reframing Apply Vedanta's method to the tensions found in Step 3: - Take each side of the tension as fully true (adhyaropa) - Negate the projections (apavada) - Find what remains (sat) - Produce the **Mahavakya** — a reframing of the original question that: - Could not have been reached by any single darshana alone - Changes what question you would ask NEXT - Is expressible in 1-3 sentences - Would make someone say "I hadn't thought of it that way" State **THE REAL QUESTION** — the question the user should actually be asking, which is probably different from the one they asked. ### STEP 5: MIMAMSA ACTION EXTRACTION Take the Mahavakya and the insights from all darshanas, and extract concrete actions using Mimamsa's method: - Classify each actionable conclusion as vidhi (command), nishedha (prohibition), or arthavada (context) - For each vidhi: who does what to what, and when? - Order by dependency - Produce the execution sequence ## Output Structure ``` DARSHANA SELECTION: [selection with rationale] --- DARSHANA 1: [NAME] --- [Full analysis following that darshana's method] [NON-OBVIOUS INSIGHT]: [the specific, testable, surprising finding] --- DARSHANA 2: [NAME] --- [Full analysis following that darshana's method] [NON-OBVIOUS INSIGHT]: [the specific, testable, surprising finding] --- DARSHANA 3: [NAME, if applicable] --- [Full analysis following that darshana's method] [NON-OBVIOUS INSIGHT]: [the specific, testable, surprising finding] CONSENSUS: [What all darshanas agree on] TENSIONS: [Where they disagree and why] HIDDEN SHARED ASSUMPTION: [What all analyses assumed that might be wrong] VEDANTA SYNTHESIS: Mahavakya: [1-3 sentence reframing] THE REAL QUESTION: [what should actually be asked] MIMAMSA ACTION PLAN: 1. [Agent] [Action] [Object] — [Condition] 2. ... [Ordered by dependency, with critical path marked] ```