@azure-id/orc 1.9.0 → 1.9.2

package/bin/onboarding-content.js

@@ -36,8 +36,8 @@ const SECTIONS = [
       "init merges two guards into .claude/settings.json (non-destructively):",
       "  • a PreToolUse effort hard-block (keeps /orc on a capable tier)",
       "  • a statusline model warning (❌ ORC WILL DEGRADE on the wrong tier)",
-      "Run your MAIN session on Opus 4.8 high (or Opus 5 / Fable 5 at medium+) —",
-      "subagents can't exceed it; the top [90,100] band needs an Opus 5 session.",
+      "Run your MAIN session on Opus 4.8 high (or Opus 5.5 / Fable 5 at medium+) —",
+      "subagents can't exceed it; the top [90,100] band needs an Opus 5.5 session.",
     ],
   },
   {
@@ -54,7 +54,7 @@ const SECTIONS = [
       "",
       "Session-tier acceptance (the statusline verdict):",
       "  ✔ ORC-ready     Opus 4.8 high            (the baseline)",
-      "  🚀 ORC-boosted  Opus 4.8 xhigh/max, or   Opus 5 / Fable 5 medium…max",
+      "  🚀 ORC-boosted  Opus 4.8 xhigh/max, or   Opus 5.5 / Fable 5 medium…max",
       "  ✖ WILL DEGRADE  anything below           (wrong model or sub-baseline effort)",
       "",
       "Which one do I type? Start from what you HAVE, not from the lane names:",
@@ -153,7 +153,7 @@ const SECTIONS = [
     title: "⑧ Troubleshooting",
     lines: [
       "  \"/orc blocked — required effort not met\"",
-      "     → switch to high effort (or Opus 5 / Fable 5 at medium+). xhigh/max also pass.",
+      "     → switch to high effort (or Opus 5.5 / Fable 5 at medium+). xhigh/max also pass.",
       "  statusline says \"WILL DEGRADE\"",
       "     → wrong model/effort, or a usage window ≥90%. Move to Opus 4.8 high or Opus 5.",
       "  \"wiki: UNREGISTERED\"",

package/bin/pricing.json

@@ -1,200 +1,207 @@
-{
-  "_readme": [
-    "Dated price table for `orc budget`. Point `orc config set budget_price_table <path>`",
-    "at your own copy to override it. Rates are USD per MILLION tokens.",
-    "",
-    "Four rates per model, never one: cache reads are usually the LARGEST token",
-    "count in a run and the CHEAPEST per token, so a blended number hides",
-    "whichever component is about to bite you. cache_write/cache_read are the",
-    "multipliers this vendor publishes against `input` (1.25x / 0.1x) written out",
-    "as absolute rates so nothing has to remember the ratio.",
-    "",
-    "`as_of` is load-bearing: a table older than 90 days makes every dollar",
-    "figure print a staleness warning. Bump the date only when you have checked",
-    "the published rates.",
-    "",
-    "`providers` (v0.50.0) is the same four-rate shape for a NON-CLAUDE endpoint",
-    "reached through `orc extra`. Every `models` map SHIPS EMPTY, on purpose, and",
-    "that is not an oversight: between this package's cutoff and any release the",
-    "lineup moves (DeepSeek V3.2 -> V4, Kimi K2 -> K2.7, GLM 4.6 -> 5.2), several",
-    "of these vendors price by PEAK/OFF-PEAK window or by tier, one is",
-    "subscription-only, and one is a passthrough with a surcharge. A shipped",
-    "figure that is wrong by 2x is worse than no figure, because a wrong figure",
-    "gets believed. So ORC ships the MECHANISM and no dollar figure without a",
-    "rate you put there yourself.",
-    "",
-    "Fill one in with `orc extra rates` — it lists every provider/model pair your",
-    "own traces have actually used, prints the JSON to paste, and says which ones",
-    "it has no rate for. A pair with no rate reports `usd: null` forever, never a",
-    "guess and never a family fallback."
-  ],
-  "as_of": "2026-08-01",
-  "currency": "USD",
-  "unit": "per_million_tokens",
-  "server_tools": {
-    "web_search_requests": 0.01,
-    "web_fetch_requests": 0.0
-  },
-  "models": {
-    "claude-opus-5": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-opus-4-8": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-opus-4-7": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-fable-5": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-sonnet-5": {
-      "input": 3,
-      "cache_write": 3.75,
-      "cache_read": 0.3,
-      "output": 15
-    },
-    "claude-sonnet-4-6": {
-      "input": 3,
-      "cache_write": 3.75,
-      "cache_read": 0.3,
-      "output": 15
-    },
-    "claude-haiku-4-5": {
-      "input": 1,
-      "cache_write": 1.25,
-      "cache_read": 0.1,
-      "output": 5
-    }
-  },
-  "families": {
-    "opus": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "fable": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "sonnet": {
-      "input": 3,
-      "cache_write": 3.75,
-      "cache_read": 0.3,
-      "output": 15
-    },
-    "haiku": {
-      "input": 1,
-      "cache_write": 1.25,
-      "cache_read": 0.1,
-      "output": 5
-    }
-  },
-  "_plans_readme": [
-    "Quota view. A Pro/Max user does not pay per token — they burn a 5-hour",
-    "session window and a weekly ceiling. These are WEIGHTED-token capacities",
-    "(cache reads counted at their 0.1x weight), because that is the unit the",
-    "forecast can actually produce. They are approximations of a published",
-    "allowance, not a metered balance: the percentages are a magnitude, and the",
-    "renderer says so."
-  ],
-  "plans": {
-    "pro": {
-      "label": "Pro",
-      "window_weighted_tokens": 900000,
-      "weekly_weighted_tokens": 9000000
-    },
-    "max5": {
-      "label": "Max 5x",
-      "window_weighted_tokens": 4500000,
-      "weekly_weighted_tokens": 45000000
-    },
-    "max20": {
-      "label": "Max 20x",
-      "window_weighted_tokens": 18000000,
-      "weekly_weighted_tokens": 180000000
-    }
-  },
-  "context_windows": {
-    "claude-opus-5": 200000,
-    "claude-opus-4-8": 200000,
-    "claude-opus-4-7": 200000,
-    "claude-fable-5": 200000,
-    "claude-sonnet-5": 200000,
-    "claude-sonnet-4-6": 200000,
-    "claude-haiku-4-5": 200000
-  },
-  "providers": {
-    "_note": "Rates are USD per MILLION tokens, four per model, never one. `models` maps a provider's own model id EXACTLY as the endpoint reports it.",
-    "deepseek": {
-      "label": "DeepSeek",
-      "caveat": "Prices differ by PEAK and OFF-PEAK window, so one number cannot be right for both. Record the window you actually run in.",
-      "models": {}
-    },
-    "zai": {
-      "label": "Z.ai / GLM",
-      "caveat": "The current flagship is sold inside a Coding Plan subscription rather than per token, which a per-token table cannot express. Leave it empty and read your invoice.",
-      "models": {}
-    },
-    "moonshot": {
-      "label": "Moonshot / Kimi",
-      "caveat": null,
-      "models": {}
-    },
-    "minimax": {
-      "label": "MiniMax",
-      "caveat": null,
-      "models": {}
-    },
-    "qwen": {
-      "label": "Alibaba / Qwen",
-      "caveat": "Tiered by request size, so a single rate is an average rather than a price.",
-      "models": {}
-    },
-    "mimo": {
-      "label": "Xiaomi / MiMo",
-      "caveat": null,
-      "models": {}
-    },
-    "stepfun": {
-      "label": "StepFun",
-      "caveat": null,
-      "models": {}
-    },
-    "siliconflow": {
-      "label": "SiliconFlow",
-      "caveat": null,
-      "models": {}
-    },
-    "openrouter": {
-      "label": "OpenRouter",
-      "caveat": "A PASSTHROUGH: the rate is the upstream provider's, plus a surcharge on credit purchase. It also reroutes between providers by default while preserving the model id, so the model id alone does not identify who you paid.",
-      "models": {}
-    },
-    "ollama": {
-      "label": "Ollama (local)",
-      "caveat": "Local inference has no per-token price. Zero here would be true of the API bill and false of the electricity and the hardware, so it is left empty rather than asserted as free.",
-      "models": {}
-    },
-    "custom": {
-      "label": "Custom endpoint",
-      "caveat": null,
-      "models": {}
-    }
-  }
-}
+{
+  "_readme": [
+    "Dated price table for `orc budget`. Point `orc config set budget_price_table <path>`",
+    "at your own copy to override it. Rates are USD per MILLION tokens.",
+    "",
+    "Four rates per model, never one: cache reads are usually the LARGEST token",
+    "count in a run and the CHEAPEST per token, so a blended number hides",
+    "whichever component is about to bite you. cache_write/cache_read are the",
+    "multipliers this vendor publishes against `input` (1.25x / 0.1x) written out",
+    "as absolute rates so nothing has to remember the ratio.",
+    "",
+    "`as_of` is load-bearing: a table older than 90 days makes every dollar",
+    "figure print a staleness warning. Bump the date only when you have checked",
+    "the published rates.",
+    "",
+    "`providers` (v0.50.0) is the same four-rate shape for a NON-CLAUDE endpoint",
+    "reached through `orc extra`. Every `models` map SHIPS EMPTY, on purpose, and",
+    "that is not an oversight: between this package's cutoff and any release the",
+    "lineup moves (DeepSeek V3.2 -> V4, Kimi K2 -> K2.7, GLM 4.6 -> 5.2), several",
+    "of these vendors price by PEAK/OFF-PEAK window or by tier, one is",
+    "subscription-only, and one is a passthrough with a surcharge. A shipped",
+    "figure that is wrong by 2x is worse than no figure, because a wrong figure",
+    "gets believed. So ORC ships the MECHANISM and no dollar figure without a",
+    "rate you put there yourself.",
+    "",
+    "Fill one in with `orc extra rates` — it lists every provider/model pair your",
+    "own traces have actually used, prints the JSON to paste, and says which ones",
+    "it has no rate for. A pair with no rate reports `usd: null` forever, never a",
+    "guess and never a family fallback."
+  ],
+  "as_of": "2026-08-01",
+  "currency": "USD",
+  "unit": "per_million_tokens",
+  "server_tools": {
+    "web_search_requests": 0.01,
+    "web_fetch_requests": 0.0
+  },
+  "models": {
+    "claude-opus-5-5": {
+      "input": 4,
+      "cache_write": 5,
+      "cache_read": 0.2,
+      "output": 20
+    },
+    "claude-opus-5": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-opus-4-8": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-opus-4-7": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-fable-5": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-sonnet-5": {
+      "input": 3,
+      "cache_write": 3.75,
+      "cache_read": 0.3,
+      "output": 15
+    },
+    "claude-sonnet-4-6": {
+      "input": 3,
+      "cache_write": 3.75,
+      "cache_read": 0.3,
+      "output": 15
+    },
+    "claude-haiku-4-5": {
+      "input": 1,
+      "cache_write": 1.25,
+      "cache_read": 0.1,
+      "output": 5
+    }
+  },
+  "families": {
+    "opus": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "fable": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "sonnet": {
+      "input": 3,
+      "cache_write": 3.75,
+      "cache_read": 0.3,
+      "output": 15
+    },
+    "haiku": {
+      "input": 1,
+      "cache_write": 1.25,
+      "cache_read": 0.1,
+      "output": 5
+    }
+  },
+  "_plans_readme": [
+    "Quota view. A Pro/Max user does not pay per token — they burn a 5-hour",
+    "session window and a weekly ceiling. These are WEIGHTED-token capacities",
+    "(cache reads counted at their 0.1x weight), because that is the unit the",
+    "forecast can actually produce. They are approximations of a published",
+    "allowance, not a metered balance: the percentages are a magnitude, and the",
+    "renderer says so."
+  ],
+  "plans": {
+    "pro": {
+      "label": "Pro",
+      "window_weighted_tokens": 900000,
+      "weekly_weighted_tokens": 9000000
+    },
+    "max5": {
+      "label": "Max 5x",
+      "window_weighted_tokens": 4500000,
+      "weekly_weighted_tokens": 45000000
+    },
+    "max20": {
+      "label": "Max 20x",
+      "window_weighted_tokens": 18000000,
+      "weekly_weighted_tokens": 180000000
+    }
+  },
+  "context_windows": {
+    "claude-opus-5-5": 200000,
+    "claude-opus-5": 200000,
+    "claude-opus-4-8": 200000,
+    "claude-opus-4-7": 200000,
+    "claude-fable-5": 200000,
+    "claude-sonnet-5": 200000,
+    "claude-sonnet-4-6": 200000,
+    "claude-haiku-4-5": 200000
+  },
+  "providers": {
+    "_note": "Rates are USD per MILLION tokens, four per model, never one. `models` maps a provider's own model id EXACTLY as the endpoint reports it.",
+    "deepseek": {
+      "label": "DeepSeek",
+      "caveat": "Prices differ by PEAK and OFF-PEAK window, so one number cannot be right for both. Record the window you actually run in.",
+      "models": {}
+    },
+    "zai": {
+      "label": "Z.ai / GLM",
+      "caveat": "The current flagship is sold inside a Coding Plan subscription rather than per token, which a per-token table cannot express. Leave it empty and read your invoice.",
+      "models": {}
+    },
+    "moonshot": {
+      "label": "Moonshot / Kimi",
+      "caveat": null,
+      "models": {}
+    },
+    "minimax": {
+      "label": "MiniMax",
+      "caveat": null,
+      "models": {}
+    },
+    "qwen": {
+      "label": "Alibaba / Qwen",
+      "caveat": "Tiered by request size, so a single rate is an average rather than a price.",
+      "models": {}
+    },
+    "mimo": {
+      "label": "Xiaomi / MiMo",
+      "caveat": null,
+      "models": {}
+    },
+    "stepfun": {
+      "label": "StepFun",
+      "caveat": null,
+      "models": {}
+    },
+    "siliconflow": {
+      "label": "SiliconFlow",
+      "caveat": null,
+      "models": {}
+    },
+    "openrouter": {
+      "label": "OpenRouter",
+      "caveat": "A PASSTHROUGH: the rate is the upstream provider's, plus a surcharge on credit purchase. It also reroutes between providers by default while preserving the model id, so the model id alone does not identify who you paid.",
+      "models": {}
+    },
+    "ollama": {
+      "label": "Ollama (local)",
+      "caveat": "Local inference has no per-token price. Zero here would be true of the API bill and false of the electricity and the hardware, so it is left empty rather than asserted as free.",
+      "models": {}
+    },
+    "custom": {
+      "label": "Custom endpoint",
+      "caveat": null,
+      "models": {}
+    }
+  }
+}

package/bin/verify-contracts.js

@@ -1813,7 +1813,7 @@ const CONTRACTS = [
     binFiles: ["bin/cli.js"],
     files: [
       "agents/MODEL-MAPPING.md",
-      // v1.8.0 — the graph noter has no Opus 5 variant; under the mode a lane
+      // v1.8.0 — the graph noter has no Opus 5.5 variant; under the mode a lane
       // skips notes. Both the agent and the canonical contract say so.
       "agents/orc-graph-noter-sonnet-4-6-med.md",
       "skills/_shared/code-graph.md",
@@ -2741,6 +2741,75 @@ const CONTRACTS = [
       "skills/orc-fast/SKILL.md",
     ],
   },
+  // ── v1.9.1 W3 — the brief answer, and the lines the CLI now computes ──
+  {
+    // Every `--json` read a lane makes carries it, and the CLI is the only
+    // thing that can answer it. A rename that reached one side and not the
+    // other would leave every lane asking for a flag nothing knows.
+    name: "graph brief answer (v1.9.1 — --brief on every --json read a lane makes)",
+    token: "--brief",
+    files: [
+      "agents/orc-executor-haiku-4-5.md",
+      "agents/orc-executor-opus-4-7-high.md",
+      "agents/orc-executor-opus-4-7-med.md",
+      "agents/orc-executor-opus-4-8-high.md",
+      "agents/orc-executor-opus-5-high.md",
+      "agents/orc-executor-opus-5-low.md",
+      "agents/orc-executor-opus-5-med.md",
+      "agents/orc-executor-sonnet-4-6-high.md",
+      "agents/orc-executor-sonnet-4-6-med.md",
+      "agents/orc-executor-sonnet-5-high.md",
+      "agents/orc-recon-opus-5-low.md",
+      "agents/orc-recon-sonnet-4-6-med.md",
+      "skills/_shared/code-graph.md",
+      "skills/_shared/phases/execution.md",
+      "skills/_shared/phases/planning.md",
+      "skills/_shared/phases/preflight.md",
+      // review.md names it to say where it is deliberately NOT used.
+      "skills/_shared/phases/review.md",
+      "skills/_shared/phases/ship.md",
+      // trace.md names the call each trace verb is copied from.
+      "skills/_shared/phases/trace.md",
+      "skills/_shared/read-ladder.md",
+      "skills/orc/SKILL.md",
+      "skills/orc-fast/SKILL.md",
+      "skills/orc-mini/SKILL.md",
+      "skills/orc-mini/references/complexity.md",
+      "skills/orc-quick/SKILL.md",
+      "skills/orc-quick/references/gh-mode.md",
+      "skills/orc-quick/references/look.md",
+    ],
+    binFiles: ["bin/cli.js"],
+  },
+  {
+    // The blast radius is the CLI's line now. A lane that assembled it by hand
+    // would drift from the CLI's grammar the first time either side moved.
+    name: "graph blast line (v1.9.1 — changes computes the blast radius, the lane prints it)",
+    token: "blast_line",
+    files: ["skills/orc-mini/SKILL.md", "skills/orc-quick/SKILL.md"],
+    binFiles: ["bin/graph-signals.js"],
+  },
+  {
+    // Mini's one call. The flag, the prose that explains it and the engine that
+    // answers it move together or not at all.
+    name: "graph complexity read (v1.9.1 — impact --complexity, one call, one line)",
+    token: "--complexity",
+    files: [
+      "skills/_shared/code-graph.md",
+      "skills/_shared/phases/trace.md",
+      "skills/orc-mini/SKILL.md",
+      "skills/orc-mini/references/complexity.md",
+    ],
+    binFiles: ["bin/cli.js"],
+  },
+  {
+    // The four thresholds live in ONE place. The prose quotes them by NAME, so
+    // a rename that reaches only the engine fails here.
+    name: "graph complexity thresholds (v1.9.1 — the constants the prose quotes)",
+    token: "COMPLEXITY_CALLER_FILES",
+    files: ["skills/orc-mini/references/complexity.md"],
+    binFiles: ["bin/graph-signals.js"],
+  },
   // ── v1.8.2 W3 — the delivery flags (D1 · D2 · D3 · D5) ────────────────
   {
     // The slice call. A rename must reach the contract, the phase file that
@@ -2773,6 +2842,43 @@ const CONTRACTS = [
     ],
     binFiles: ["bin/cli.js"],
   },
+  // ── v1.9.1 W5 — delivery where the paid reads are (R1 · R4) ─────────────
+  {
+    // The call sites WITH the card. The recon agents ask for it, the ladder
+    // and the contract say what it may never print, and the CLI answers it.
+    name: "graph callers source (v1.9.1 — ctx --callers-source, six lines per confident call site)",
+    token: "--callers-source",
+    files: [
+      "agents/orc-recon-opus-5-low.md",
+      "agents/orc-recon-sonnet-4-6-med.md",
+      "skills/_shared/code-graph.md",
+      "skills/_shared/read-ladder.md",
+    ],
+    binFiles: ["bin/cli.js", "bin/graph-query.js"],
+  },
+  {
+    // Where a language server is asked. Every reader that is told to use it
+    // names the field, and the engine is the only thing that computes it.
+    name: "graph LSP anchor (v1.9.1 — lsp_at on a symbol card: file, line, 1-based character)",
+    token: "lsp_at",
+    files: [
+      "agents/orc-executor-haiku-4-5.md",
+      "agents/orc-executor-opus-4-7-high.md",
+      "agents/orc-executor-opus-4-7-med.md",
+      "agents/orc-executor-opus-4-8-high.md",
+      "agents/orc-executor-opus-5-high.md",
+      "agents/orc-executor-opus-5-low.md",
+      "agents/orc-executor-opus-5-med.md",
+      "agents/orc-executor-sonnet-4-6-high.md",
+      "agents/orc-executor-sonnet-4-6-med.md",
+      "agents/orc-executor-sonnet-5-high.md",
+      "agents/orc-recon-opus-5-low.md",
+      "agents/orc-recon-sonnet-4-6-med.md",
+      "skills/_shared/code-graph.md",
+      "skills/_shared/read-ladder.md",
+    ],
+    binFiles: ["bin/graph-query.js"],
+  },
   // ── v1.9.0 — the lean lanes reproduce before they fix ───────────────────
   {
     // The slice field that makes a defect entry show the bug RED first. It is
@@ -2943,7 +3049,7 @@ const CONTRACTS = [
   {
     // The only notes writer, dispatched BY NAME. An unnamed dispatch cannot
     // enforce the Sonnet 4.6 pin and is invisible to the trace hook.
-    name: "graph noter (v1.8.0 — dispatched by name, no Opus 5 variant)",
+    name: "graph noter (v1.8.0 — dispatched by name, no Opus 5.5 variant)",
     token: "orc-graph-noter-sonnet-4-6-med",
     files: [
       "agents/MODEL-MAPPING.md",
@@ -4562,6 +4668,15 @@ for (const b of BUDGETS) {
   //    one. design-05 §6.3: a call one lane makes belongs to that lane.
   for (const [id, c] of Object.entries(CALLS)) {
     const want = [...(measured[prefixOf(c.cmd)] || new Set())].sort();
+    // v1.9.1: a row with `lanes: []` is a USER command — catalogued so
+    // `orc lane calls --all` documents it, and asserted the OTHER way round:
+    // no lane may name it. `orc graph audit` is the first one. The assertion is
+    // stronger than the >= 2 rule, not an exception to it.
+    if (Array.isArray(c.lanes) && c.lanes.length === 0) {
+      if (want.length)
+        errs.push(`${id}: a user command, and ${want.join(", ")} name it — remove the call, or give the row its lanes`);
+      continue;
+    }
     if (want.length < 2)
       errs.push(
         `${id}: only ${want.length} lane names it — a call one lane makes belongs to that lane and is not catalogued`