@azure-id/orc 2.0.2 → 2.1.0

package/mock-run/orc-explain.md

@@ -1,86 +1,86 @@
-# Mock run — `/orc-explain`
-
-> ORC said something you did not understand. Type this and it says it again,
-> properly.
-
----
-
-## 1. What it does
-
-It re-explains the **last thing ORC said** (or an earlier one, if you point at
-it). It adds the background the message assumed, and defines every ORC-only
-word in your project's own terms.
-
-It is not "make it shorter". A summary of something you did not understand is
-the same thing you did not understand.
-
-Nothing about the run changes: no re-plan, no re-score, no undo, no new
-question. When it finishes, you are exactly where you were.
-
----
-
-## 2. The run
-
-ORC has just printed a scoring block. Rina reads it and blinks.
-
-```
-T1 3·new-surface·stateful·new-tests·fan0/3·low = 6+18+16+8+9 = 57
-   → orc-executor-sonnet-5-high (keystone)
-```
-
-```
-> /orc-explain
-```
-
-```
-The point: task T1 is medium-hard, so I am sending it to Sonnet 5 at high
-effort, and it must finish before three other tasks can start.
-
-The background that line assumed
-  Before building, I give each task a number from 0 to 100. The number is
-  arithmetic, not opinion: the planner (which read your files) reports a few
-  properties of the task, and I add them up with a published formula. The
-  number then picks the model from a fixed table. That is why you can argue
-  with it — the inputs are all visible.
-
-Every word in that line, in this project's terms
-  facet          one property of the task the planner measured
-  3              the task touches 3 files (`breadth`)
-  new-surface    api/orders/note_service.ts does not exist yet — this task
-                 creates it (`novelty`)
-  stateful       it holds data between calls, not just if/else (`logic`)
-  new-tests      tests must be written, not just updated (`test_surface`)
-  fan0/3         it depends on 0 tasks; 3 tasks depend on IT. That is why I
-                 called it a keystone, and why it runs alone in wave 1.
-  low            the planner's own confidence that the task is uncertain
-  57             the total. The table maps [55,65) to Sonnet 5 at high effort.
-  band           one row of that table
-  wave           a group of tasks that can run at the same time because no
-                 two of them write the same file
-
-What this means for you
-  Nothing to decide. This was me showing my work before dispatching. If you
-  think 57 is wrong — say so, and I will re-score with your correction.
-
-Back to the run. Waiting for your "go" on the wave plan.
-```
-
----
-
-## 3. What to notice
-
-- **It explains, it never re-does.** No agent is dispatched, no file is
-  written, no gate changes.
-- **It defines words in *your* project's terms**, using `wiki/orc-orientation.md`
-  when a wiki exists, so the words match the rest of your docs.
-- **The skill itself is deliberately tiny.** A long skill that says "be clear"
-  makes a wordy model, because the model copies the length it sees.
-- **This lane writes no trace on purpose.** It is the only lane that does not,
-  and `orc stats --help` says so — a stated blind spot, not an oversight.
-
----
-
-## 4. Related
-
-- The words themselves: `orc onboarding lanes`
-- The whole walkthrough with no GitHub needed: `orc onboarding`
+# Mock run — `/orc-explain`
+
+> ORC said something you did not understand. Type this and it says it again,
+> properly.
+
+---
+
+## 1. What it does
+
+It re-explains the **last thing ORC said** (or an earlier one, if you point at
+it). It adds the background the message assumed, and defines every ORC-only
+word in your project's own terms.
+
+It is not "make it shorter". A summary of something you did not understand is
+the same thing you did not understand.
+
+Nothing about the run changes: no re-plan, no re-score, no undo, no new
+question. When it finishes, you are exactly where you were.
+
+---
+
+## 2. The run
+
+ORC has just printed a scoring block. Rina reads it and blinks.
+
+```
+T1 3·new-surface·stateful·new-tests·fan0/3·low = 6+18+16+8+9 = 57
+   → orc-executor-opus-5-low (keystone)
+```
+
+```
+> /orc-explain
+```
+
+```
+The point: task T1 is medium-hard, so I am sending it to Opus 5.5 at low
+effort, and it must finish before three other tasks can start.
+
+The background that line assumed
+  Before building, I give each task a number from 0 to 100. The number is
+  arithmetic, not opinion: the planner (which read your files) reports a few
+  properties of the task, and I add them up with a published formula. The
+  number then picks the model from a fixed table. That is why you can argue
+  with it — the inputs are all visible.
+
+Every word in that line, in this project's terms
+  facet          one property of the task the planner measured
+  3              the task touches 3 files (`breadth`)
+  new-surface    api/orders/note_service.ts does not exist yet — this task
+                 creates it (`novelty`)
+  stateful       it holds data between calls, not just if/else (`logic`)
+  new-tests      tests must be written, not just updated (`test_surface`)
+  fan0/3         it depends on 0 tasks; 3 tasks depend on IT. That is why I
+                 called it a keystone, and why it runs alone in wave 1.
+  low            the planner's own confidence that the task is uncertain
+  57             the total. The table maps [41,90) to Opus 5.5 at low effort.
+  band           one row of that table
+  wave           a group of tasks that can run at the same time because no
+                 two of them write the same file
+
+What this means for you
+  Nothing to decide. This was me showing my work before dispatching. If you
+  think 57 is wrong — say so, and I will re-score with your correction.
+
+Back to the run. Waiting for your "go" on the wave plan.
+```
+
+---
+
+## 3. What to notice
+
+- **It explains, it never re-does.** No agent is dispatched, no file is
+  written, no gate changes.
+- **It defines words in *your* project's terms**, using `wiki/orc-orientation.md`
+  when a wiki exists, so the words match the rest of your docs.
+- **The skill itself is deliberately tiny.** A long skill that says "be clear"
+  makes a wordy model, because the model copies the length it sees.
+- **This lane writes no trace on purpose.** It is the only lane that does not,
+  and `orc stats --help` says so — a stated blind spot, not an oversight.
+
+---
+
+## 4. Related
+
+- The words themselves: `orc onboarding lanes`
+- The whole walkthrough with no GitHub needed: `orc onboarding`