crewforth 3.5.0-rc.1 → 3.5.0-rc.2

package/kit/skills/spec-planning/SKILL.md

@@ -32,16 +32,23 @@ commit: <HEAD when the plan was written>
 ## Constraints
 - <a value, limit or name copied word for word from the request or spec>
 ## Tasks (order)
-1. <task> — satisfies: AC-1 — size: <S/M/L> — dependency: <none/#n> — risk: <low/medium/high>
+Progress: tick a task [x] when its check passes, mark the one in work [~]. /crew-todo shows these boxes.
+### Phase 1 — <name>                                              (phases only when the work has stages)
+1. [ ] <task> — satisfies: AC-1 — size: <S/M/L> — dependency: <none/#n> — risk: <low/medium/high>
    Interfaces: consumes <name: type> · produces <name: type>      (only when a later task uses its output)
    Current code: <a few lines, verbatim, at the place it changes>  (optional)
-2. …
+2. [ ] …
 — checkpoint: build and tests green, the criteria so far re-read —  (after every 2-3 tasks)
+### Phase 2 — <name>
+3. [ ] …
 ## Not covered by tests
 - <an input or case no test reaches> → task <n>                    (five at most)
 ## Assumptions / Open questions
 - ...
 ```
+- **Every task carries a box, and the box is the record.** `[ ]` open, `[x]` done (its check passed, not "written"),
+  `[~]` the one in work. Tick it in the same step that finishes the task; the `Progress:` line keeps that rule in the
+  plan, where whoever works it reads it. `/crew-todo` draws the boxes under the status line; it reads, never writes.
 - **No timeline and no estimate in time.** A size (S/M/L) says how much, never when.
 - **No code beyond signatures.** A plan names functions and types; the bodies are the tasks' work. The only code a
   plan quotes is the current code a task changes, word for word.

package/kit/studio/web/agent-spend.js

@@ -0,0 +1,32 @@
+// What each kind of agent spent, for the summary box: under "Agents", one line per agent type.
+//
+// The numbers are the graph's: every agent node carries its own share of the session's tokens (`usage`), counted
+// once per response whichever file repeated it. This file only adds them up by type, with the same rule the
+// graph's groups use (graph-plan.js `sumOf`): an agent with no usage read adds nothing, and a cost is a dash when
+// any member's is, so a sum never passes for the whole.
+
+import { sumOf } from './graph-plan.js';
+import { fmtCount, fmtCost } from './usage.js';
+import { typeOf, roleName } from './names.js';
+
+/**
+ * [role, "new tokens · ~cost"] pairs, the most-spent first: the shape `summaryRows` opens under a row.
+ * A type that ran more than once says how many times. Empty when no agent has a usage to add.
+ */
+export function agentSpendRows(nodes) {
+  const byType = new Map();
+  for (const n of nodes ?? []) {
+    if (n?.kind !== 'agent' || !n.usage) continue;
+    const type = typeOf(n);
+    if (!byType.has(type)) byType.set(type, []);
+    byType.get(type).push(n);
+  }
+  return [...byType]
+    .map(([type, members]) => ({ type, runs: members.length, sum: sumOf(members) }))
+    .filter((g) => g.sum.fresh || g.sum.cacheRead)
+    .sort((a, b) => b.sum.fresh - a.sum.fresh || (a.type < b.type ? -1 : 1))
+    .map((g) => [
+      g.runs > 1 ? `${roleName(g.type)} ×${g.runs}` : roleName(g.type),
+      `${fmtCount(g.sum.fresh)} · ${fmtCost(g.sum.cost)}`,
+    ]);
+}