[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7APWBTV4S2YRDP6T0R0FERZ":3,"skill-same:01M7APWBTV4S2YRDP6T0R0FERZ":50},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":33,"ownerDescription":34,"updatedAt":35,"install":36,"resources":49},"01M7APWBTV4S2YRDP6T0R0FERZ","forge-self-consistency",null,"Governs N perspective-diverse, mutually-blind reasoning traces with structural-agreement judge-merge for gated high-stakes FORGE decisions (G2\u002FTRANCE). Use instead of forge-deliberation when Deep depth AND a stakes flag both hold (Framing score 8–9 with irreversible\u002Fsystem-wide blast radius), or on explicit opt-in. Do NOT use for standard-tier decisions — self-consistency is never the default.","Rynaro\u002Fmagicite","Rynaro","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F5271543?v=4","01M70CFYFMB316W23THVQAQ42M","Henrique A. Lavezzo","01M7APN2KCVWBNY727T57KR2ZJ",2,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002F.eidolons\u002Fforge\u002Fskills\u002Fself-consistency\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: forge-self-consistency\ndescription: Governs N perspective-diverse, mutually-blind reasoning traces with structural-agreement judge-merge for gated high-stakes FORGE decisions (G2\u002FTRANCE). Use instead of forge-deliberation when Deep depth AND a stakes flag both hold (Framing score 8–9 with irreversible\u002Fsystem-wide blast radius), or on explicit opt-in. Do NOT use for standard-tier decisions — self-consistency is never the default.\nmetadata:\n  methodology: FORGE\n  phase: R-variant\n---\n\n# Self-Consistency Mode\n\nLoaded **instead of** `skills\u002Fdeliberation\u002FSKILL.md` when a G2 \u002F TRANCE self-consistency\nescalation fires. Governs how the Reasoner runs **N perspective-diverse,\nmutually-blind reasoning traces** over a frozen Frame+Observe inventory and\nmerges them on **structural agreement**, not on verbalized confidence.\n\n## When to use\n\nLoad this skill **instead of** `forge-deliberation` when BOTH a complexity flag AND a stakes flag hold: Deep depth (Framing Step-4 total 8–9) with an irreversible or system-wide blast radius. Also loads on explicit opt-in (\"run self-consistency\", \"G2\", \"N=5 on this\"). Use N=3 for standard high-stakes runs and N=5 for the worst-case quadrant (irreversible AND system-wide). Do not enter this mode for standard-depth decisions; do not combine with `forge-deliberation` in the same Reason phase.\n\nThis mode is **gated, never default** (see \"Gate\" below). Standard-tier FORGE\nstays single-trace (`skills\u002Fdeliberation\u002FSKILL.md`, 1–3 passes). Self-consistency is\nthe TRANCE form the trance-matrix grants FORGE: *\"N=3 (or N=5 high-stakes)\nsampled traces with majority-vote \u002F judge-merge (G2)\"* — operationalized here.\n\n---\n\n## Gate — when this mode fires (and when it must not)\n\nSelf-consistency fires ONLY when **both** a complexity flag and a stakes flag\nhold, OR on explicit opt-in, OR when the roster declares a weak-host degraded\nmode (see item 3 below):\n\n1. **Deep depth** — Framing Step-4 total 8–9 (high Ambiguity AND Irreversibility\n   OR system-wide Blast radius). A Deep score on its own is necessary but the\n   stakes flag (irreversible \u002F system-wide) must also be live. This is the\n   trance-matrix C6 auto-trigger requirement: a complexity flag AND a stakes flag.\n2. **Explicit opt-in** — the user or the parent orchestrator names the mode\n   (\"run self-consistency\", \"G2\", \"N=5 on this\").\n3. **Weak or undeclared host** — the roster declares FORGE\n   `degraded_mode: sample-select` (`roster\u002Frouting.yaml`, routing-1.1): on a\n   weak or undeclared host this mode REPLACES single-trace self-red-teaming\n   (the Pass-3 Deep red-team in `skills\u002Fdeliberation\u002FSKILL.md`) rather than only\n   gating on Deep + stakes — the roster data itself is the trigger, because\n   sampling+selection is more reliable than unbounded single-trace\n   self-correction on a weak host.\n\nIf none of the above hold, **do not enter this mode** — run standard single-trace FORGE.\nForcing G2 universally violates the opt-in contract (FORGE is never a mandatory\ncritique gate). This mode adds *parallelism*, not a fresh budget: it composes\nwith — does not extend — the existing 3-pass \u002F 1-REFORGE cap.\n\n**No tool access is gained at TRANCE.** FORGE is reasoning-only at every tier.\nThe independent-verifier hook below is a *handoff recommendation*, not an\nexecuted call.\n\n---\n\n## Step 1 — Freeze the shared inventory (fan-out precondition)\n\nBefore fan-out, the Frame and Observe outputs are **frozen**: the decision\nquestion, decision type, constraint table, success criteria, and the evidence\ninventory (with H\u002FM\u002FL reliability tiers). Every trace reasons from this **same**\nfrozen inventory. Traces differ in *reasoning stance*, never in the evidence\nthey are given — a trace that invents new evidence is invalid and is discarded.\n\nPick N:\n\n| N | When |\n|---|------|\n| **N=3** | Standard high-stakes (Deep depth + stakes flag). |\n| **N=5** | Irreversible AND system-wide blast radius (the worst-case quadrant). |\n\nN is **fixed before fan-out** and bounded — exactly N traces, **no re-sampling\nbeyond N**. If the merge is below floor, you emit `[DISPUTED]`; you do not draw\nmore traces.\n\n---\n\n## Step 2 — Fan-out: N perspective-diverse, mutually-blind traces\n\nDiversity is across **reasoning stance**, not model (FORGE has no model control;\nintra-model sample diversity already beats naive cross-model debate). Each trace\nis assigned a distinct adversarial lens from this fixed persona table, in order,\nuntil N is reached:\n\n| # | Persona | Lens — the question this trace asks first |\n|---|---------|-------------------------------------------|\n| 1 | **Evidence-maximizing** | Which hypothesis is best supported by the H-reliability evidence, weighing alignment above all? |\n| 2 | **Pre-mortem \u002F failure-first** | Assume each hypothesis was chosen and failed — which fails *least catastrophically*? |\n| 3 | **Constraint-relaxation** | If exactly one soft constraint could be relaxed, which hypothesis becomes clearly best, and is that relaxation acceptable? |\n| 4 | **Inversion \u002F steelman-the-rejected** | Build the strongest case for the hypothesis the obvious reading rejects — does it survive? |\n| 5 | **Simplest-viable-default** | What is the lowest-complexity, most-reversible option that still satisfies all hard constraints? |\n\nFor N=3 use personas 1–3. For N=5 use 1–5.\n\n**Mutual blindness is mandatory.** No trace may read another trace's verdict,\nscore, or selected hypothesis before the merge. This prevents self-conditioning\n(a trace anchoring on an earlier trace's answer collapses the diversity the mode\nexists to create — the lazy-agent failure).\n\nTwo equivalent execution forms:\n\n- **Parent-dispatched subagents (mechanically stronger).** The orchestrator\n  dispatches N clean-context FORGE subagents, each given the frozen inventory\n  plus one persona, with no shared scratchpad. Independence is enforced by\n  separate contexts. Prefer this form when available.\n- **Sequential-blind in one context (fallback).** Emit N reasoning passes one\n  after another, each opening from the frozen inventory under its assigned\n  persona, and **do not reference any prior pass's verdict** while reasoning the\n  current one. Only the merge step (Step 3) may read all N.\n\nEach trace produces the standard hypothesis-level work (≥3 hypotheses, the four\nstress tests, a selected hypothesis, and any `[RISK]` \u002F `[REVERSAL-CONDITION]`\nmarkers it surfaces). A trace MAY state a per-trace confidence, but that number\nis **not** used by the merge.\n\n---\n\n## Step 3 — Aggregation: structural-agreement judge-merge\n\nThe merge is a **single deterministic tally**, not another round of\ndeliberation. There is exactly **one** merge pass — no merge loop.\n\n1. **Tally hypothesis selection.** For each trace, record which hypothesis it\n   selected. Identify the **modal hypothesis** (the one selected by the most\n   traces).\n2. **Compute the structural-agreement score:**\n\n   ```\n   structural_agreement = (traces selecting the modal hypothesis) \u002F N\n   ```\n\n3. **Cross-reference recurring markers.** Collect every `[REVERSAL-CONDITION]`\n   and `[RISK]` that **≥2 traces surfaced independently**. Independently-recurring\n   reversal conditions are **high-trust signals** — they were found by\n   differently-biased reasoners, so they are unlikely to be persona artefacts.\n4. **Decide against the 60% consensus floor** (trance-matrix FORGE consensus\n   floor):\n\n   | structural_agreement | Result |\n   |----------------------|--------|\n   | **≥ 60%** | **PASS** — emit one merged `[VERDICT]` on the modal hypothesis. |\n   | **\u003C 60%** | **FAIL** — emit `[DISPUTED]` (see Step 4). |\n\n**Merged confidence anchors on structural agreement, NOT on averaged verbalized\nconfidence.** The merged verdict's confidence IS the `structural_agreement`\nscore (optionally tightened upward only by the count of independently-recurring\nreversal conditions). Do **not** average the per-trace confidence numbers:\nverbalized confidence is systematically overconfident and a poor diagnostic of\ncorrectness, so it is excluded from the trust anchor. The mechanical signal —\n*how many independently-biased traces converged* — is the calibrated one.\n\nA PASS verdict lists: the modal hypothesis, the structural-agreement score as\nits confidence, and the convergent (`≥2`-trace) reversal conditions as the\nverdict's `[REVERSAL-CONDITION]` set.\n\n---\n\n## Step 4 — Disagreement: `[DISPUTED]` below the floor\n\nIf `structural_agreement \u003C 60%`, **never force-merge** into a single winner.\nEmit a `[DISPUTED]` verdict that:\n\n- enumerates the **live positions** (each hypothesis that ≥1 trace selected,\n  with which persona(s) backed it),\n- states a **reduced confidence** equal to the structural-agreement score,\n- lists which reversal conditions recurred and which were trace-local, and\n- routes a handoff: **→ human** (genuine judgment split) or, if a tiebreak is\n  feasible, **→ independent verifier** (see Step 5).\n\nThis mirrors the single-trace tied-score rule: when the top positions are\ngenuinely close, the verdict acknowledges ambiguity rather than manufacturing a\nwinner.\n\n---\n\n## Step 5 — Independent-verifier hook (opt-in ceiling-breaker)\n\nSelf-consistency on a single reasoner saturates around a known accuracy ceiling;\nthe lever that pushes past it is an **independent \u002F cross-model verifier**.\nFORGE is tool-less (P0) and cannot dispatch a second model itself, so this is a\n**documented handoff, not an executed call**:\n\n- **When to offer it.** A near-floor PASS (structural_agreement just above 60%),\n  a `[DISPUTED]` FAIL on a high-stakes\u002Firreversible decision, or explicit opt-in.\n- **What to emit.** An `[ACTION]` \u002F handoff recommending that the parent\n  orchestrator (or nexus `eidolons run`) re-derive the verdict via an\n  **independent FORGE instance or a cross-model verifier**, seeded with the\n  **same frozen inventory** but **not** shown this run's merged verdict. Name it\n  explicitly as `→ independent-verifier` in the handoff section.\n- **What NOT to do.** Do not call a tool. Do not start a debate loop between\n  traces. The verifier is a separate, blind re-derivation orchestrated outside\n  FORGE — that independence is exactly what makes it a ceiling-breaker.\n\n---\n\n## Bounds & honesty\n\n- **Exactly N traces, one merge pass.** No re-sampling, no merge loop. Composes\n  with the 3-pass \u002F 1-REFORGE cap; does not extend it.\n- **Diversity is tuned, not naive.** Fixed adversarial personas are the *tuned*\n  form of multi-trace reasoning; naive same-prompt sampling or unbounded debate\n  is excluded (see `DESIGN-RATIONALE.md §9`).\n- **Confidence is structural.** It reflects cross-trace agreement, never an\n  averaged self-grade. This is a methodology-layer estimate, not a benchmarked\n  accuracy figure.\n\n---\n\n*Reasoner — Self-Consistency Skill*\n","# Self-Consistency Mode\n\nLoaded **instead of** `skills\u002Fdeliberation\u002FSKILL.md` when a G2 \u002F TRANCE self-consistency\nescalation fires. Governs how the Reasoner runs **N perspective-diverse,\nmutually-blind reasoning traces** over a frozen Frame+Observe inventory and\nmerges them on **structural agreement**, not on verbalized confidence.\n\n## When to use\n\nLoad this skill **instead of** `forge-deliberation` when BOTH a complexity flag AND a stakes flag hold: Deep depth (Framing Step-4 total 8–9) with an irreversible or system-wide blast radius. Also loads on explicit opt-in (\"run self-consistency\", \"G2\", \"N=5 on this\"). Use N=3 for standard high-stakes runs and N=5 for the worst-case quadrant (irreversible AND system-wide). Do not enter this mode for standard-depth decisions; do not combine with `forge-deliberation` in the same Reason phase.\n\nThis mode is **gated, never default** (see \"Gate\" below). Standard-tier FORGE\nstays single-trace (`skills\u002Fdeliberation\u002FSKILL.md`, 1–3 passes). Self-consistency is\nthe TRANCE form the trance-matrix grants FORGE: *\"N=3 (or N=5 high-stakes)\nsampled traces with majority-vote \u002F judge-merge (G2)\"* — operationalized here.\n\n---\n\n## Gate — when this mode fires (and when it must not)\n\nSelf-consistency fires ONLY when **both** a complexity flag and a stakes flag\nhold, OR on explicit opt-in, OR when the roster declares a weak-host degraded\nmode (see item 3 below):\n\n1. **Deep depth** — Framing Step-4 total 8–9 (high Ambiguity AND Irreversibility\n   OR system-wide Blast radius). A Deep score on its own is necessary but the\n   stakes flag (irreversible \u002F system-wide) must also be live. This is the\n   trance-matrix C6 auto-trigger requirement: a complexity flag AND a stakes flag.\n2. **Explicit opt-in** — the user or the parent orchestrator names the mode\n   (\"run self-consistency\", \"G2\", \"N=5 on this\").\n3. **Weak or undeclared host** — the roster declares FORGE\n   `degraded_mode: sample-select` (`roster\u002Frouting.yaml`, routing-1.1): on a\n   weak or undeclared host this mode REPLACES single-trace self-red-teaming\n   (the Pass-3 Deep red-team in `skills\u002Fdeliberation\u002FSKILL.md`) rather than only\n   gating on Deep + stakes — the roster data itself is the trigger, because\n   sampling+selection is more reliable than unbounded single-trace\n   self-correction on a weak host.\n\nIf none of the above hold, **do not enter this mode** — run standard single-trace FORGE.\nForcing G2 universally violates the opt-in contract (FORGE is never a mandatory\ncritique gate). This mode adds *parallelism*, not a fresh budget: it composes\nwith — does not extend — the existing 3-pass \u002F 1-REFORGE cap.\n\n**No tool access is gained at TRANCE.** FORGE is reasoning-only at every tier.\nThe independent-verifier hook below is a *handoff recommendation*, not an\nexecuted call.\n\n---\n\n## Step 1 — Freeze the shared inventory (fan-out precondition)\n\nBefore fan-out, the Frame and Observe outputs are **frozen**: the decision\nquestion, decision type, constraint table, success criteria, and the evidence\ninventory (with H\u002FM\u002FL reliability tiers). Every trace reasons from this **same**\nfrozen inventory. Traces differ in *reasoning stance*, never in the evidence\nthey are given — a trace that invents new evidence is invalid and is discarded.\n\nPick N:\n\n| N | When |\n|---|------|\n| **N=3** | Standard high-stakes (Deep depth + stakes flag). |\n| **N=5** | Irreversible AND system-wide blast radius (the worst-case quadrant). |\n\nN is **fixed before fan-out** and bounded — exactly N traces, **no re-sampling\nbeyond N**. If the merge is below floor, you emit `[DISPUTED]`; you do not draw\nmore traces.\n\n---\n\n## Step 2 — Fan-out: N perspective-diverse, mutually-blind traces\n\nDiversity is across **reasoning stance**, not model (FORGE has no model control;\nintra-model sample diversity already beats naive cross-model debate). Each trace\nis assigned a distinct adversarial lens from this fixed persona table, in order,\nuntil N is reached:\n\n| # | Persona | Lens — the question this trace asks first |\n|---|---------|-------------------------------------------|\n| 1 | **Evidence-maximizing** | Which hypothesis is best supported by the H-reliability evidence, weighing alignment above all? |\n| 2 | **Pre-mortem \u002F failure-first** | Assume each hypothesis was chosen and failed — which fails *least catastrophically*? |\n| 3 | **Constraint-relaxation** | If exactly one soft constraint could be relaxed, which hypothesis becomes clearly best, and is that relaxation acceptable? |\n| 4 | **Inversion \u002F steelman-the-rejected** | Build the strongest case for the hypothesis the obvious reading rejects — does it survive? |\n| 5 | **Simplest-viable-default** | What is the lowest-complexity, most-reversible option that still satisfies all hard constraints? |\n\nFor N=3 use personas 1–3. For N=5 use 1–5.\n\n**Mutual blindness is mandatory.** No trace may read another trace's verdict,\nscore, or selected hypothesis before the merge. This prevents self-conditioning\n(a trace anchoring on an earlier trace's answer collapses the diversity the mode\nexists to create — the lazy-agent failure).\n\nTwo equivalent execution forms:\n\n- **Parent-dispatched subagents (mechanically stronger).** The orchestrator\n  dispatches N clean-context FORGE subagents, each given the frozen inventory\n  plus one persona, with no shared scratchpad. Independence is enforced by\n  separate contexts. Prefer this form when available.\n- **Sequential-blind in one context (fallback).** Emit N reasoning passes one\n  after another, each opening from the frozen inventory under its assigned\n  persona, and **do not reference any prior pass's verdict** while reasoning the\n  current one. Only the merge step (Step 3) may read all N.\n\nEach trace produces the standard hypothesis-level work (≥3 hypotheses, the four\nstress tests, a selected hypothesis, and any `[RISK]` \u002F `[REVERSAL-CONDITION]`\nmarkers it surfaces). A trace MAY state a per-trace confidence, but that number\nis **not** used by the merge.\n\n---\n\n## Step 3 — Aggregation: structural-agreement judge-merge\n\nThe merge is a **single deterministic tally**, not another round of\ndeliberation. There is exactly **one** merge pass — no merge loop.\n\n1. **Tally hypothesis selection.** For each trace, record which hypothesis it\n   selected. Identify the **modal hypothesis** (the one selected by the most\n   traces).\n2. **Compute the structural-agreement score:**\n\n   ```\n   structural_agreement = (traces selecting the modal hypothesis) \u002F N\n   ```\n\n3. **Cross-reference recurring markers.** Collect every `[REVERSAL-CONDITION]`\n   and `[RISK]` that **≥2 traces surfaced independently**. Independently-recurring\n   reversal conditions are **high-trust signals** — they were found by\n   differently-biased reasoners, so they are unlikely to be persona artefacts.\n4. **Decide against the 60% consensus floor** (trance-matrix FORGE consensus\n   floor):\n\n   | structural_agreement | Result |\n   |----------------------|--------|\n   | **≥ 60%** | **PASS** — emit one merged `[VERDICT]` on the modal hypothesis. |\n   | **\u003C 60%** | **FAIL** — emit `[DISPUTED]` (see Step 4). |\n\n**Merged confidence anchors on structural agreement, NOT on averaged verbalized\nconfidence.** The merged verdict's confidence IS the `structural_agreement`\nscore (optionally tightened upward only by the count of independently-recurring\nreversal conditions). Do **not** average the per-trace confidence numbers:\nverbalized confidence is systematically overconfident and a poor diagnostic of\ncorrectness, so it is excluded from the trust anchor. The mechanical signal —\n*how many independently-biased traces converged* — is the calibrated one.\n\nA PASS verdict lists: the modal hypothesis, the structural-agreement score as\nits confidence, and the convergent (`≥2`-trace) reversal conditions as the\nverdict's `[REVERSAL-CONDITION]` set.\n\n---\n\n## Step 4 — Disagreement: `[DISPUTED]` below the floor\n\nIf `structural_agreement \u003C 60%`, **never force-merge** into a single winner.\nEmit a `[DISPUTED]` verdict that:\n\n- enumerates the **live positions** (each hypothesis that ≥1 trace selected,\n  with which persona(s) backed it),\n- states a **reduced confidence** equal to the structural-agreement score,\n- lists which reversal conditions recurred and which were trace-local, and\n- routes a handoff: **→ human** (genuine judgment split) or, if a tiebreak is\n  feasible, **→ independent verifier** (see Step 5).\n\nThis mirrors the single-trace tied-score rule: when the top positions are\ngenuinely close, the verdict acknowledges ambiguity rather than manufacturing a\nwinner.\n\n---\n\n## Step 5 — Independent-verifier hook (opt-in ceiling-breaker)\n\nSelf-consistency on a single reasoner saturates around a known accuracy ceiling;\nthe lever that pushes past it is an **independent \u002F cross-model verifier**.\nFORGE is tool-less (P0) and cannot dispatch a second model itself, so this is a\n**documented handoff, not an executed call**:\n\n- **When to offer it.** A near-floor PASS (structural_agreement just above 60%),\n  a `[DISPUTED]` FAIL on a high-stakes\u002Firreversible decision, or explicit opt-in.\n- **What to emit.** An `[ACTION]` \u002F handoff recommending that the parent\n  orchestrator (or nexus `eidolons run`) re-derive the verdict via an\n  **independent FORGE instance or a cross-model verifier**, seeded with the\n  **same frozen inventory** but **not** shown this run's merged verdict. Name it\n  explicitly as `→ independent-verifier` in the handoff section.\n- **What NOT to do.** Do not call a tool. Do not start a debate loop between\n  traces. The verifier is a separate, blind re-derivation orchestrated outside\n  FORGE — that independence is exactly what makes it a ceiling-breaker.\n\n---\n\n## Bounds & honesty\n\n- **Exactly N traces, one merge pass.** No re-sampling, no merge loop. Composes\n  with the 3-pass \u002F 1-REFORGE cap; does not extend it.\n- **Diversity is tuned, not naive.** Fixed adversarial personas are the *tuned*\n  form of multi-trace reasoning; naive same-prompt sampling or unbounded debate\n  is excluded (see `DESIGN-RATIONALE.md §9`).\n- **Confidence is structural.** It reflects cross-trace agreement, never an\n  averaged self-grade. This is a methodology-layer estimate, not a benchmarked\n  accuracy figure.\n\n---\n\n*Reasoner — Self-Consistency Skill*\n",{"name":5,"metadata":24,"description":7},{"phase":25,"methodology":26},"R-variant","FORGE","main","38c3a82583abf9806de764f89ac0099ad690933a",".eidolons\u002Fforge\u002Fskills\u002Fself-consistency",0,"https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite","https:\u002F\u002Fgithub.com\u002FRynaro","https:\u002F\u002Fhlavezzo.me","A Code Alchemist with a passion for weaving tales and inscribing transmutation circles in pure code.","2026-10-14 02:35:37",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add Rynaro\u002Fmagicite --skill forge-self-consistency",{"tool":42,"command":43},"bunx","bunx skills add Rynaro\u002Fmagicite --skill forge-self-consistency",{"tool":45,"command":46},"pnpm","pnpm dlx skills add Rynaro\u002Fmagicite --skill forge-self-consistency","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002F.eidolons\u002Fforge\u002Fskills\u002Fself-consistency\u002FSKILL.md",true,[],[51,58,65,72,79,86],{"skillNo":52,"name":53,"nameZh":6,"description":54,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":55,"categories":56,"keywords":57,"publishedAt":19},"01M7AQ4GHJQVV8TJBJ8T1V7ZQ9","wine-dxvk-cache-clear","Clear a stale DXVK shader cache that is causing graphical corruption in a\nWine or Proton game. Use when: a game shows corrupted textures or shader\nflicker that started after a driver or DXVK version change.","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Ftoy-registry\u002Fskills\u002Fwine-dxvk-cache-clear\u002FSKILL.md",[],[],{"skillNo":59,"name":60,"nameZh":6,"description":61,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":19},"01M7AQ4CY1NHSMY1HCZ91387YY","steam-download-region-fix","Fix a Steam client stuck at very low download speed by changing its\ndownload region. Use when: Steam downloads are far below the local\ninternet speed for every game, not just one.","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Ftoy-registry\u002Fskills\u002Fsteam-download-region-fix\u002FSKILL.md",[],[],{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":19},"01M7AQ49P7Y0CDRSYMPVABX80K","proton-battleye-eac-toggle","Toggle BattlEye or Easy Anti-Cheat runtime support for a Steam Proton game.\nUse when: a multiplayer game refuses to launch with an anti-cheat error under Proton.","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Ftoy-registry\u002Fskills\u002Fproton-battleye-eac-toggle\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":19},"01M7AQ45326FAP3X4Q86FK1X4N","vivi-verify-incoming","Load when reading any upstream artefact handed off to Vivi that carries a sibling .envelope.json. BLOCKING per ECL §6.2.2: the orchestrator MUST have verified the envelope's SHA-256 (eidolons verify-envelope --block \u002F eidolons run --verify) and recorded a verify_pass before dispatch. If no verify_pass exists for the message_id, or a verify_fail is present, REFUSE to process the payload and hand back to the orchestrator. Symmetric receiver gate — every Eidolon enforces it identically.","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002F.eidolons\u002Fvivi\u002Fskills\u002Fverify-incoming\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":19},"01M7AQ402861P0P3C2TEGFGMA1","vivi-parallel-tracks","Load ONLY under TRANCE authorization when the Plan phase yields N independent, non-overlapping implementation tracks (disjoint file sets). Operationalizes the TRANCE G4 form — parallel multi-track implementation in mandatory git-worktree isolation, max 5 clean-context tracks, a per-track verifier cascade, a non-fungible per-track ≤3 reflection budget, explicit stop conditions, and a single-threaded merge\u002Faggregation step. Absent TRANCE gating, Vivi runs exactly as today: single-track A→P→I→V→Δ\u002FR.","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002F.eidolons\u002Fvivi\u002Fskills\u002Fparallel-tracks\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":19},"01M7AQ3RH9BZF9ZMTC7VG5P7PB","vivi-methodology","Full Vivi cycle reference — A(nalyze) → P(lan) → I(mplement) → V(erify) → Δ(Delta)\u002FR(eflect). Use when the task is a non-trivial feature implementation in a brownfield codebase and you need the complete methodology reference: complexity routing, evidence-grounded planning rules, test-anchoring requirements, and failure-escalation thresholds.","https:\u002F\u002Fgithub.com\u002FRynaro\u002Fmagicite\u002Fblob\u002Fmain\u002F.eidolons\u002Fvivi\u002Fskills\u002Fmethodology\u002FSKILL.md",[],[]]