@azure-id/orc 2.1.0 → 2.2.0

package/templates/skills/orc/SKILL.md

@@ -24,8 +24,8 @@ version of a reference you haven't loaded this run. Detect the stack from the re
 
 Confirm you are **Opus 4.8 high**, or **Opus 5.5 / Fable 5 at medium+** (both clear
 the guard from medium up; Opus 5.5 also unlocks the [90,100] executor band). Effort
-is hard-blocked by the `orc-effort-guard.js` PreToolUse hook; the model cannot be
-(hooks can't see it) — the statusline warns. On a weaker tier **STOP immediately**
+is hard-blocked by `orc-effort-guard.js` (a Skill call: the live effort; a typed `/orc`:
+the status line's reading, never a guess); the model cannot be — the statusline warns. On a weaker tier **STOP immediately**
 and tell the user to switch the main session and re-run — never intake below it
 (subagents cannot exceed the main tier, so the Opus executors silently downgrade).
 
@@ -37,7 +37,7 @@ themselves are this lane's own and stay here.
 
 1. **You NEVER implement. You coordinate.** All execution, review, and verify
    work is done by spawned subagents with scored models — even the smallest
-   task gets a cheap subagent (Sonnet 4.6 medium), never you.
+   task gets a cheap subagent (the lowest band, Sonnet 5 low), never you.
 2. **Disk is truth; conversation is a cache.** On any resume or suspected
    compaction: re-read `state-of-play.md` then the checkpoint BEFORE acting.
 3. **All run artifacts go in `.claude/orc/run/{run-slug}/`** (config `run_dir`,
@@ -100,8 +100,8 @@ session on Opus or the Opus pins silently fall back (the original "wrong model"
 
 ## Lane contract (`../_shared/lane-contract.md` — read it ONLY when a call exits ≠ 0)
 
-- **Calls:** `orc lane calls orc --json` names every call and its exit codes.
-  **An exit code is an ANSWER where it says so, not a failure.** Make no other call.
+- **Calls:** `orc lane calls orc --json` names shared calls and exit codes.
+  **An exit code is an ANSWER where it says so, not a failure.** Own calls are here; make no other.
 - **Config:** `orc lane config orc --json`. Obey `effective`, print every line
   in `announce[]` VERBATIM at preflight, and honour `stops[]` before wave 1.
   Never merge `.claude/orc.config.yaml` yourself (`../_shared/config-precedence.md`).
@@ -120,7 +120,7 @@ in the SAME step; store `trace_path`.
 **Under `ultra_mode` the lane segment is `ultra`, not `orc`** (`run-ultra-<slug>-…`)
 — the filename IS the per-lane data, so an ultra run named `orc` is counted as a
 plain `/orc` run forever, hiding the costliest lane in every usage report.
-**Phase close = `orc trace write --packet -`** (exit ≠ 0 → `orc-trace-writer-haiku-4-5`).
+**Phase close = `orc trace write --packet -`** (exit ≠ 0 → `orc-trace-writer-haiku-5-high`).
 The first packet is solo — it repairs a hook-bootstrap filename.
 `DISPATCH`/`VERIFY` models are derived from the agent NAME and checked against
 each return's `actual_model`/`actual_effort` — surface any ⛔ DOWNGRADE to the
@@ -179,7 +179,7 @@ them; they are never dispatched as subagents).
 - Phase 5.5 → `../_shared/phases/security-checklist.md`; 6.5 → `subskills/orc-testgen/`
 - Phase 8 → `subskills/orc-pr/SKILL.md` (template `subskills/orc-pr/pr.md`);
   stack gate → `subskills/orc-pr/stack-gate.md` + `_shared/pr-templates.md`
-- Code graph cache, NEVER skipped (§0 of the contract) — `orc graph status --if-enabled --heal --json` at preflight BEFORE the first dispatch (builds or updates it) · ONE `orc graph ctx <declared files>` call per slice · `orc graph map` FIRST at planning (orientation — before any Glob, before `impact`) then `orc graph impact` + `orc graph cochange`, `orc graph changes` at review · `orc graph coverage` before you trust a card's silence · `orc graph update` + `orc graph notes` at every wave close and ship · every `--json` read carries `--brief` · print each JSON `line`, copy each `trace` VERBATIM (CLI-resolved) → `../_shared/code-graph.md`
+- Code graph — the four touchpoints (`../_shared/code-graph.md` §0), every call with `--if-enabled --json --brief`: (1) consult at preflight — print `probes["graph-status"].line`; `orc graph status --heal` only when `heal_needed`; exit 3 (`off` or `semi`) → no other graph call · (2) orient at planning — `orc graph map --focus=<3–6 words> --budget 800` ONCE, then `orc graph impact` + `orc graph cochange` on the candidate files · (3) anchors — one `orc graph ctx --for-slice <declared files> --focus=<words> [--refs=<task.precedent>] --anchors` per writing dispatch (a request that names one symbol: `ctx <symbol> --anchors` gives the card and the anchors in one call); its `card` goes into the slice VERBATIM as `graph_anchors`, with its `read all: …` line and its code block — never shorten it; With `graph_anchors` in the slice, list at most ONE wiki page path (the one about the feature's intent), not 1–3. ( none for one named file with no outside change) · (4) affected tests — `orc graph changes --files=<actual files>` at review. `orc graph update --notes-pending` only when `update_step` is `lane`. No lane runs the gain meter. Print each `line`; copy each `trace` VERBATIM
 - Schemas (you own; pass slices only): `schemas/intent-spec.md`,
   `schemas/planning-output.md`, `schemas/checkpoint.md`
 - Worked example (orient only — never execute from it) → `examples/full-run-mock.md`

package/templates/skills/orc/config.md

@@ -43,8 +43,8 @@ EFFORT, which is why the old four Opus rows are two. **Every band from score 41
 needs an Opus 5.5 MAIN session** or it silently falls back to the session model
 (the tier-honesty rule reports the downgrade) — that is the cost of this table.
 
-**Seven executors are named by NO band** — `orc-executor-haiku-4-5`,
-`orc-executor-sonnet-4-6-med`, `orc-executor-sonnet-4-6-high`,
+**Six executors are named by NO band** — `orc-executor-haiku-5-high`,
+`orc-executor-sonnet-4-6-high`,
 `orc-executor-opus-4-7-med`, `orc-executor-opus-4-7-high`,
 `orc-executor-opus-4-8-high` and `orc-executor-opus-5-high`. They still ship,
 reachable through `rubric_bands_override`, `orc diy`'s `fixed_executor` and

package/templates/skills/orc/examples/full-run-mock.md

@@ -22,17 +22,17 @@ O: [6 tasks with declared_files incl. tests + per-file grounding attestations +
 O: Refined: medium, partially dependent. Recommend PARALLEL waves. Confirm? U: yes. Batch pause? U: every 2 waves.
 O: Facet scoring (planner facets → formula → band; facet gate: breadth + fan recomputed, risk cites checked):
    T1 3·new-surface·stateful·new-tests·fan0/3·low = 6+18+16+8+9 = 57 → sonnet-5-high (keystone — every later task imports its enum)
-   T2 3·imitate·branching·new-tests·fan1/0·low = 6+8+8+8+5 = 35 → sonnet-4-6-med
-   T3 3·imitate·branching·update-existing·fan1/0·low = 6+8+8+4+5 = 31 → sonnet-4-6-med
-   T4 2·imitate·branching·update-existing·fan1/0·low = 6+8+8+4+5 = 31 → sonnet-4-6-med
-   T5 1·mechanical·none·update-existing·fan0/0·low = 2+0+0+4 = 6 → haiku-4-5
-   T6 1·mechanical·none·none·fan0/0·low = 2+0+0+0 = 2 → haiku-4-5
+   T2 3·imitate·branching·new-tests·fan1/0·low = 6+8+8+8+5 = 35 → sonnet-5-high
+   T3 3·imitate·branching·update-existing·fan1/0·low = 6+8+8+4+5 = 31 → sonnet-5-high
+   T4 2·imitate·branching·update-existing·fan1/0·low = 6+8+8+4+5 = 31 → sonnet-5-high
+   T5 1·mechanical·none·update-existing·fan0/0·low = 2+0+0+4 = 6 → haiku-5-high
+   T6 1·mechanical·none·none·fan0/0·low = 2+0+0+0 = 2 → haiku-5-high
    Plan: 6 tasks, 3 waves W1{T1} W2{T2,T3,T4} W3{T5,T6}, all spawned. Anticipated escalations? U: none, go.
 
 ## Phase 3 — Execution (ALL spawned; orchestrator coordinates only)
 O: [checkpoint + state-of-play written into run/merchant-notifications/ BEFORE dispatch]
 ▶ W1: spawn Agent-A(T1, sonnet-5-high) — milestones 40%→100% — return validated ✓ (evidence: `yarn jest models/` exit 0, tail quoted; unmet[] empty) — log: DECISION type enum.
-▶ W2: spawn 3 agents (sonnet-4-6-med each).
+▶ W2: spawn 3 agents (sonnet-5-high each).
    Agent-C needs_context (1/2): "needs T1's enum interface" → O adjudicates in-scope → re-slices → resumes. (User never bothered.)
    Agent-C user escalation: "digest send time?" → U: 07:00 fixed → ANSWER broadcast to log.
    Returns validated ✓ · collision audit clean.
@@ -46,7 +46,7 @@ O: [checkpoint + state-of-play written into run/merchant-notifications/ BEFORE d
 
 U: [next day, FRESH session, pastes block]
 O: [state-of-play → checkpoint → re-anchored] "Resuming: merchant notifications (in-app+digest+prefs). Still correct?" U: yes.
-▶ W3: spawn T5+T6 (haiku-4-5) → done. All 6 complete.
+▶ W3: spawn T5+T6 (haiku-5-high) → done. All 6 complete.
 
 ## Phase 5 — Review (spawned worker, Opus 5.5 medium)
 O: Code pattern? U: none. Findings (P0–P3 ladder, each P0–P2 anchored file:line + verbatim quote):

package/templates/skills/orc/references/effort-and-mode.md

@@ -1,222 +1,222 @@
-# Reference — Effort, Dispatch Style, and Task Scoring
-
-**Subagents ALWAYS do the work — the orchestrator never implements** (hard
-rule 1). The orchestrator is Opus 4.8 high: it doing implementation is the most
-costly way possible, and it burns the context that must survive the whole run.
-Spawning keeps orchestrator context lean so runs last longer before any pause.
-
-Two INDEPENDENT axes. Never conflate them:
-- **Run-level effort** (low/medium/high) → picks the **dispatch style**
-  (sequential vs parallel), which controls only **intra-wave concurrency** —
-  never WHO does the work, and never WHETHER waves exist.
-- **Per-task score** (0–100) → picks each **worker's model**. Always applies.
-
-A medium run can contain a 30-score task and a 95-score task in the same wave.
-
-## Waves always exist; dispatch style is intra-wave only
-
-**Wave computation runs for every run with ≥2 tasks, sequential included**
-(dependency layers + conflict graph + `max_wave_tasks` cap — see
-`wave-grouping.md`). Dispatch style does NOT decide whether there are waves; it
-decides how a wave's tasks fire:
-
-- **sequential** → a wave's tasks dispatch one at a time, in order; the wave
-  closes when all its tasks close;
-- **parallel** → a wave's non-conflicting tasks dispatch at once (up to
-  `max_wave_tasks`).
-
-The **wave-boundary batch pause binds to wave numbers, identically in both
-styles** — a sequential run is NOT "no waves / per-task pauses". Show the wave
-plan (wave → tasks → pause marks) to the user BEFORE wave 1 in both styles.
-
-## Run-level effort → dispatch style (recommend, user confirms)
-
-- **Low**, or heavy shared data/code → **sequential**: one scored subagent at a
-  time, in dependency order (still grouped into waves). Even a single trivial
-  task = one cheap subagent (typically Sonnet 4.6 medium), never the
-  orchestrator itself.
-- **Medium** → sequential by default; **parallel** intra-wave if 3+ genuinely
-  independent areas.
-- **High** with independent areas → **parallel** waves; consider worktrees for
-  isolation (merge at end).
-
-Always RECOMMEND with a one-line why, and let the user pick. Tell them the
-agent/task/wave counts, each task's scored model, and where batch pauses fall
-BEFORE dispatching.
-
-## Per-task scoring — facet-scored, arithmetic, two-writer
-
-The score is **not judged** from a task title. The **planner** — the one party
-that globbed and read every declared file — emits per-task `facets` (facts). The
-**orchestrator** — who never read the code — computes the number arithmetically
-and audits the facts. No vibes anywhere. Every task is scored, including tiny ones.
-
-**1. Planner-emitted facets** (`planning-output.md` `facets` block, filled during
-grounding — zero extra passes):
-
-| Facet | Values |
-|---|---|
-| `breadth` | `len(declared_files)` — computed, not judged |
-| `novelty` | mechanical · imitate · new-surface · novel-algorithm |
-| `logic` | none · branching · stateful · algorithmic |
-| `test_surface` | none · update-existing · new-tests |
-| `risk` | `[]` or `[{class, cite}]` — class ∈ auth·money·migration·security·concurrency·data-integrity; **each entry MUST cite the file/requirement** |
-| `uncertainty` | low · medium · high (+ one-line reason if not low) |
-
-`fan_in`/`fan_out` are NOT emitted — the orchestrator computes them from
-`depends_on` (forward = fan_in, reverse = fan_out).
-
-**2. Orchestrator validation gate (Phase 2, deterministic — same bounce
-mechanics as grounding):** emit `GATE facet pass|bounce`. Three checks:
-
-- **Recompute** `breadth` (= `len(declared_files)`) and `fan_in`/`fan_out` from
-  the plan itself — a mismatch bounces.
-- **MEMBERSHIP (v0.34.4):** `novelty`, `logic`, `test_surface` and `uncertainty`
-  must each be a member of their CLOSED set above, and every `risk[].class` must
-  be one of the six risk classes. A set lookup, as deterministic as the grounding
-  Glob. This is the check that was missing: the gate validated only the two
-  facets it could recompute and TRUSTED the four it could not, so a planner that
-  invented a `low|medium|high` scale produced an **arithmetically unscorable**
-  plan that passed the stated gate. There is no `N("low")` — and an orchestrator
-  applying `risk != [] -> floor 70` to prose risk strings floors EVERY task, a
-  two-band overshoot caused by a field's shape rather than by the work.
-- **Cite check:** a `risk` entry with no `cite` bounces (unchanged).
-
-A facet-vocabulary miss is a **FIELD-SHAPE bounce**: hand back the miss list and
-say *do not re-plan* — the planner corrects only the offending `facets` blocks.
-(It may also legitimately DROP a risk entry on the correction pass: implementation
-hazards like test seeding or dependency discipline are not risk CLASSES. That
-judgment is the planner's to make and is worth accepting.) Without this clause a
-shape bounce risks costing a correct plan. A plan with no `facets` at all
-(pre-v0.31.0) resumes on the legacy path — never bounced for the missing block.
-
-**3. The fixed formula (the ONLY scoring text — this replaces base+adjusters):**
-
-```
-score = B(breadth) + N(novelty) + L(logic) + T(test_surface)
-        + 5*min(fan_in,3) + 3*min(fan_out,3) + U(uncertainty)
-
-B: 1f=2   2-3f=6   4-5f=10   6+f=15
-N: mechanical=0   imitate=8   new-surface=18   novel-algorithm=30
-L: none=0   branching=8   stateful=16   algorithmic=24
-T: none=0   update-existing=4   new-tests=8
-U: low=0   medium=6   high=12
-
-risk ≠ [] → floor 70 (DERIVED from a cited risk facet, never remembered).
-clamp 0..100 → the RESOLVED table (below).
-```
-
-**Which table (highest wins):** `opus5_only: true` (the 2-band Opus-5-only
-preset: `[0,90)` low · `[90,100]` medium) → `rubric_bands_override`
-(hand-written rows) → the default 5-band table. All three are in `config.md`.
-The formula, the facets and the risk floor are IDENTICAL in every case — only
-the score→agent mapping changes. Three consequences worth stating so nobody
-re-derives them per run:
-
-- **The risk floor still applies** — it raises the SCORE, then the resolved table
-  maps it. A floored task (≥70) lands `opus-5-low` in the default table and
-  `opus-5-low` under the Opus-5-only preset too: since v1.0.0 both tables answer
-  the floor with the same agent, and only a score of 90+ moves it to `opus-5-med`.
-- **`opus5_only` FORCES** — while on, a hand-written `rubric_bands_override` is
-  ignored. It is the one selector that can shadow another; that is deliberate.
-
-Show the user the full table (task, the facet vector, the arithmetic
-`B+N+L+T+fan+U = raw`, any risk floor, final, override+reason if any, dispatched
-model) BEFORE dispatching — an un-shown number is not a scored number. **Head it
-with the RESOLVED table's name** (`5-band default` / `Opus-5-only ladder
-(opus5_only)` / `custom (rubric_bands_override)`): the same logic
-applies to the mapping as to the number.
-
-**Extra (v0.50.0, `extra_enabled`) adds a `via` column and can make the head a
-PAIR.** An Extra route row is an OVERLAY that outranks the tables above **only
-for the scores it covers**, so a run can genuinely be running two tables at once
-and the head names both (`5-band default + extra rows [0,30) [30,70)`). The
-column reads `claude` or `extra:<profile> (<engine>)` and comes from
-`orc extra resolve --json` — the formula, the facets and the risk floor are
-untouched, and **the score is computed before the routing, never after it**. Two
-consequences the table has to show rather than imply: a cited `risk[]` HOLDS THE
-TASK BACK to its Claude band whatever the row says (`extra_risk_tasks: off`), and
-the `model` for a foreign task is the FOREIGN model id, not a Claude tier.
-Canonical: `../../_shared/extra-dispatch.md`.
-
-**4. Consistency check:** two tasks whose facet vectors differ in **≤1 facet must
-land in the same band** — or the SCORE line for the outlier **cites the one
-differing facet**. This is the specific fix for sibling tasks in the same domain
-drifting into three different bands.
-
-**5. Override protocol** (unchanged): you may override the computed score. Record
-`{computed_score, override_score, reason}` in the dispatch log. An override
-without a reason is invalid.
-
-## EVERY dispatch is scored — the fix-cycle rule
-
-Fix-cycle dispatches — review-fix, verify-fix, the Phase-7 P2-batch, and any
-requeue — are scored through the **same formula**, with two floors that stop a
-fix in a risk area from silently dropping to a cheap model:
-
-- **Inherit the ORIGINAL task's `risk` facets** when the fix touches its files —
-  a fix in a risk-floor area keeps the ≥70 floor (a userID/context-key fix can
-  never dispatch below `opus-5-low` again);
-- a **P0/P1 fix never dispatches below the band of the task that produced the
-  finding.**
-
-Emit a `SCORE task=fix-<n> …` line for each (they appear in the dispatch log and
-the completion table like any task).
-
-## Worked scoring examples (facet vectors — compare facet-by-facet, not by vibes)
-
-Each row is the facet vector run through the formula. The number is arithmetic;
-the band is what matters. Compare a new task to these facet-by-facet.
-
-| Task | breadth·novelty·logic·test · fan_in/out · unc | Arithmetic | Band |
-|---|---|---|---|
-| Rename a config key across 4 files + its test | 5·mechanical·none·update-existing · 0/0 · low | 10+0+0+4 = **14** | sonnet-5-low [0,21) |
-| New CRUD endpoint following a sibling route | 3·imitate·branching·new-tests · 1/0 · low | 6+8+8+8+5 = **35** | sonnet-5-high [31,41) |
-| Bug fix across 2 files with a repro test | 3·imitate·branching·new-tests · 0/0 · medium | 6+8+8+8+6 = **36** | sonnet-5-high [31,41) |
-| Isolated component from the design system | 3·new-surface·branching·new-tests · 0/0 · low | 6+18+8+8 = **40** | sonnet-5-high [31,41) |
-| Notification model + enum other tasks consume | 3·new-surface·stateful·new-tests · 0/3 · low | 6+18+16+8+9 = **57** | opus-5-low [41,90) |
-| Service-layer refactor behind a stable interface | 5·imitate·stateful·update-existing · 0/3 · medium | 10+8+16+4+9+6 = **53** | opus-5-low [41,90) |
-| Add role check to payment-refund endpoint | 3·imitate·branching·new-tests · 0/0 · low · **risk=[auth,money]** | 30 raw → **floor 70** | opus-5-low [41,90) |
-| Migrate orders table to split-name + backfill | 6·new-surface·stateful·new-tests · 1/3 · high · **risk=[migration,data-integrity]** | 15+18+16+8+5+9+12 = 83 (floor 70) → **83** | opus-5-low [41,90) |
-
-Two disciplines the vectors encode: (1) a small diff is NOT a low score when a
-cited `risk` facet forces the floor (the refund row — 30 raw, floored to 70); (2)
-a big-looking task IS low when its facets are mechanical (the rename). The risk
-floor is always DERIVED from a cited facet — the SCORE line names it, never
-applies it silently.
-
-## Model ladder → the single score→model table
-
-The score→model mapping is NOT hardcoded here — it lives in `config.md` as ONE
-canonical 5-band table (there is no longer a narrow/wide preset). Read config at
-run start and map each task's final score through that table (or
-`rubric_bands_override`). The orchestrator dispatches the executor agent BY NAME;
-it does not request a raw model. `rubric_bands` sets only how many bands the
-rubric REPORTS (score granularity), never which table is used.
-
-The 5 bands (see config.md for the exact edges): `sonnet-5-low` [0,21) ·
-`sonnet-5-med` [21,31) · `sonnet-5-high` [31,41) · `opus-5-low` [41,90) ·
-`opus-5-med` [90,100]. Effort tiers rank
-`low < medium < high < xhigh < max`.
-
-## Fixed model assignments (not scored)
-
-- **Orchestrator (you):** Opus 4.8 high — or Opus 5.5 / Fable 5 at medium+, the
-  two models that clear the guard from medium up. Never downgrade yourself.
-- **Review** — Superpowers path: Sonnet 4.6, medium. OpenSpec/self path:
-  Opus 5.5, low (`orc-reviewer-opus-5-low`; medium until v2.1.0).
-- **Verify:** Opus 5.5, medium (`orc-verifier-opus-5-med`).
-- **Analyst:** Opus 5.5, high. **Planner:** Opus 5.5, medium. **Test author:**
-  Opus 5.5, medium. **Combiner:** Opus 5.5, high. **Ultra advisor/judge:** Opus 5.5,
-  xhigh — every core fixed role is pinned to Opus 5.5 as of v0.34.0.
-- **Merge-conflict resolver:** Opus 4.8, medium.
-
-Note: every band in the table above is a real dispatch target (sonnet-5 through
-opus-5). If a model tier is unavailable in the environment, fall back UP to the
-next capable tier (never silently substitute a different family/effort).
-
-## Dispatch log (lives in the checkpoint)
-
-Every spawn records: `{task_id, computed_score, override_score|null,
-override_reason|null, model, effort, spawned_at}`. Feeds the usage report at
-every stop.
+# Reference — Effort, Dispatch Style, and Task Scoring
+
+**Subagents ALWAYS do the work — the orchestrator never implements** (hard
+rule 1). The orchestrator is Opus 4.8 high: it doing implementation is the most
+costly way possible, and it burns the context that must survive the whole run.
+Spawning keeps orchestrator context lean so runs last longer before any pause.
+
+Two INDEPENDENT axes. Never conflate them:
+- **Run-level effort** (low/medium/high) → picks the **dispatch style**
+  (sequential vs parallel), which controls only **intra-wave concurrency** —
+  never WHO does the work, and never WHETHER waves exist.
+- **Per-task score** (0–100) → picks each **worker's model**. Always applies.
+
+A medium run can contain a 30-score task and a 95-score task in the same wave.
+
+## Waves always exist; dispatch style is intra-wave only
+
+**Wave computation runs for every run with ≥2 tasks, sequential included**
+(dependency layers + conflict graph + `max_wave_tasks` cap — see
+`wave-grouping.md`). Dispatch style does NOT decide whether there are waves; it
+decides how a wave's tasks fire:
+
+- **sequential** → a wave's tasks dispatch one at a time, in order; the wave
+  closes when all its tasks close;
+- **parallel** → a wave's non-conflicting tasks dispatch at once (up to
+  `max_wave_tasks`).
+
+The **wave-boundary batch pause binds to wave numbers, identically in both
+styles** — a sequential run is NOT "no waves / per-task pauses". Show the wave
+plan (wave → tasks → pause marks) to the user BEFORE wave 1 in both styles.
+
+## Run-level effort → dispatch style (recommend, user confirms)
+
+- **Low**, or heavy shared data/code → **sequential**: one scored subagent at a
+  time, in dependency order (still grouped into waves). Even a single trivial
+  task = one cheap subagent (typically Sonnet 5 low), never the
+  orchestrator itself.
+- **Medium** → sequential by default; **parallel** intra-wave if 3+ genuinely
+  independent areas.
+- **High** with independent areas → **parallel** waves; consider worktrees for
+  isolation (merge at end).
+
+Always RECOMMEND with a one-line why, and let the user pick. Tell them the
+agent/task/wave counts, each task's scored model, and where batch pauses fall
+BEFORE dispatching.
+
+## Per-task scoring — facet-scored, arithmetic, two-writer
+
+The score is **not judged** from a task title. The **planner** — the one party
+that globbed and read every declared file — emits per-task `facets` (facts). The
+**orchestrator** — who never read the code — computes the number arithmetically
+and audits the facts. No vibes anywhere. Every task is scored, including tiny ones.
+
+**1. Planner-emitted facets** (`planning-output.md` `facets` block, filled during
+grounding — zero extra passes):
+
+| Facet | Values |
+|---|---|
+| `breadth` | `len(declared_files)` — computed, not judged |
+| `novelty` | mechanical · imitate · new-surface · novel-algorithm |
+| `logic` | none · branching · stateful · algorithmic |
+| `test_surface` | none · update-existing · new-tests |
+| `risk` | `[]` or `[{class, cite}]` — class ∈ auth·money·migration·security·concurrency·data-integrity; **each entry MUST cite the file/requirement** |
+| `uncertainty` | low · medium · high (+ one-line reason if not low) |
+
+`fan_in`/`fan_out` are NOT emitted — the orchestrator computes them from
+`depends_on` (forward = fan_in, reverse = fan_out).
+
+**2. Orchestrator validation gate (Phase 2, deterministic — same bounce
+mechanics as grounding):** emit `GATE facet pass|bounce`. Three checks:
+
+- **Recompute** `breadth` (= `len(declared_files)`) and `fan_in`/`fan_out` from
+  the plan itself — a mismatch bounces.
+- **MEMBERSHIP (v0.34.4):** `novelty`, `logic`, `test_surface` and `uncertainty`
+  must each be a member of their CLOSED set above, and every `risk[].class` must
+  be one of the six risk classes. A set lookup, as deterministic as the grounding
+  Glob. This is the check that was missing: the gate validated only the two
+  facets it could recompute and TRUSTED the four it could not, so a planner that
+  invented a `low|medium|high` scale produced an **arithmetically unscorable**
+  plan that passed the stated gate. There is no `N("low")` — and an orchestrator
+  applying `risk != [] -> floor 70` to prose risk strings floors EVERY task, a
+  two-band overshoot caused by a field's shape rather than by the work.
+- **Cite check:** a `risk` entry with no `cite` bounces (unchanged).
+
+A facet-vocabulary miss is a **FIELD-SHAPE bounce**: hand back the miss list and
+say *do not re-plan* — the planner corrects only the offending `facets` blocks.
+(It may also legitimately DROP a risk entry on the correction pass: implementation
+hazards like test seeding or dependency discipline are not risk CLASSES. That
+judgment is the planner's to make and is worth accepting.) Without this clause a
+shape bounce risks costing a correct plan. A plan with no `facets` at all
+(pre-v0.31.0) resumes on the legacy path — never bounced for the missing block.
+
+**3. The fixed formula (the ONLY scoring text — this replaces base+adjusters):**
+
+```
+score = B(breadth) + N(novelty) + L(logic) + T(test_surface)
+        + 5*min(fan_in,3) + 3*min(fan_out,3) + U(uncertainty)
+
+B: 1f=2   2-3f=6   4-5f=10   6+f=15
+N: mechanical=0   imitate=8   new-surface=18   novel-algorithm=30
+L: none=0   branching=8   stateful=16   algorithmic=24
+T: none=0   update-existing=4   new-tests=8
+U: low=0   medium=6   high=12
+
+risk ≠ [] → floor 70 (DERIVED from a cited risk facet, never remembered).
+clamp 0..100 → the RESOLVED table (below).
+```
+
+**Which table (highest wins):** `opus5_only: true` (the 2-band Opus-5-only
+preset: `[0,90)` low · `[90,100]` medium) → `rubric_bands_override`
+(hand-written rows) → the default 5-band table. All three are in `config.md`.
+The formula, the facets and the risk floor are IDENTICAL in every case — only
+the score→agent mapping changes. Three consequences worth stating so nobody
+re-derives them per run:
+
+- **The risk floor still applies** — it raises the SCORE, then the resolved table
+  maps it. A floored task (≥70) lands `opus-5-low` in the default table and
+  `opus-5-low` under the Opus-5-only preset too: since v1.0.0 both tables answer
+  the floor with the same agent, and only a score of 90+ moves it to `opus-5-med`.
+- **`opus5_only` FORCES** — while on, a hand-written `rubric_bands_override` is
+  ignored. It is the one selector that can shadow another; that is deliberate.
+
+Show the user the full table (task, the facet vector, the arithmetic
+`B+N+L+T+fan+U = raw`, any risk floor, final, override+reason if any, dispatched
+model) BEFORE dispatching — an un-shown number is not a scored number. **Head it
+with the RESOLVED table's name** (`5-band default` / `Opus-5-only ladder
+(opus5_only)` / `custom (rubric_bands_override)`): the same logic
+applies to the mapping as to the number.
+
+**Extra (v0.50.0, `extra_enabled`) adds a `via` column and can make the head a
+PAIR.** An Extra route row is an OVERLAY that outranks the tables above **only
+for the scores it covers**, so a run can genuinely be running two tables at once
+and the head names both (`5-band default + extra rows [0,30) [30,70)`). The
+column reads `claude` or `extra:<profile> (<engine>)` and comes from
+`orc extra resolve --json` — the formula, the facets and the risk floor are
+untouched, and **the score is computed before the routing, never after it**. Two
+consequences the table has to show rather than imply: a cited `risk[]` HOLDS THE
+TASK BACK to its Claude band whatever the row says (`extra_risk_tasks: off`), and
+the `model` for a foreign task is the FOREIGN model id, not a Claude tier.
+Canonical: `../../_shared/extra-dispatch.md`.
+
+**4. Consistency check:** two tasks whose facet vectors differ in **≤1 facet must
+land in the same band** — or the SCORE line for the outlier **cites the one
+differing facet**. This is the specific fix for sibling tasks in the same domain
+drifting into three different bands.
+
+**5. Override protocol** (unchanged): you may override the computed score. Record
+`{computed_score, override_score, reason}` in the dispatch log. An override
+without a reason is invalid.
+
+## EVERY dispatch is scored — the fix-cycle rule
+
+Fix-cycle dispatches — review-fix, verify-fix, the Phase-7 P2-batch, and any
+requeue — are scored through the **same formula**, with two floors that stop a
+fix in a risk area from silently dropping to a cheap model:
+
+- **Inherit the ORIGINAL task's `risk` facets** when the fix touches its files —
+  a fix in a risk-floor area keeps the ≥70 floor (a userID/context-key fix can
+  never dispatch below `opus-5-low` again);
+- a **P0/P1 fix never dispatches below the band of the task that produced the
+  finding.**
+
+Emit a `SCORE task=fix-<n> …` line for each (they appear in the dispatch log and
+the completion table like any task).
+
+## Worked scoring examples (facet vectors — compare facet-by-facet, not by vibes)
+
+Each row is the facet vector run through the formula. The number is arithmetic;
+the band is what matters. Compare a new task to these facet-by-facet.
+
+| Task | breadth·novelty·logic·test · fan_in/out · unc | Arithmetic | Band |
+|---|---|---|---|
+| Rename a config key across 4 files + its test | 5·mechanical·none·update-existing · 0/0 · low | 10+0+0+4 = **14** | sonnet-5-low [0,21) |
+| New CRUD endpoint following a sibling route | 3·imitate·branching·new-tests · 1/0 · low | 6+8+8+8+5 = **35** | sonnet-5-high [31,41) |
+| Bug fix across 2 files with a repro test | 3·imitate·branching·new-tests · 0/0 · medium | 6+8+8+8+6 = **36** | sonnet-5-high [31,41) |
+| Isolated component from the design system | 3·new-surface·branching·new-tests · 0/0 · low | 6+18+8+8 = **40** | sonnet-5-high [31,41) |
+| Notification model + enum other tasks consume | 3·new-surface·stateful·new-tests · 0/3 · low | 6+18+16+8+9 = **57** | opus-5-low [41,90) |
+| Service-layer refactor behind a stable interface | 5·imitate·stateful·update-existing · 0/3 · medium | 10+8+16+4+9+6 = **53** | opus-5-low [41,90) |
+| Add role check to payment-refund endpoint | 3·imitate·branching·new-tests · 0/0 · low · **risk=[auth,money]** | 30 raw → **floor 70** | opus-5-low [41,90) |
+| Migrate orders table to split-name + backfill | 6·new-surface·stateful·new-tests · 1/3 · high · **risk=[migration,data-integrity]** | 15+18+16+8+5+9+12 = 83 (floor 70) → **83** | opus-5-low [41,90) |
+
+Two disciplines the vectors encode: (1) a small diff is NOT a low score when a
+cited `risk` facet forces the floor (the refund row — 30 raw, floored to 70); (2)
+a big-looking task IS low when its facets are mechanical (the rename). The risk
+floor is always DERIVED from a cited facet — the SCORE line names it, never
+applies it silently.
+
+## Model ladder → the single score→model table
+
+The score→model mapping is NOT hardcoded here — it lives in `config.md` as ONE
+canonical 5-band table (there is no longer a narrow/wide preset). Read config at
+run start and map each task's final score through that table (or
+`rubric_bands_override`). The orchestrator dispatches the executor agent BY NAME;
+it does not request a raw model. `rubric_bands` sets only how many bands the
+rubric REPORTS (score granularity), never which table is used.
+
+The 5 bands (see config.md for the exact edges): `sonnet-5-low` [0,21) ·
+`sonnet-5-med` [21,31) · `sonnet-5-high` [31,41) · `opus-5-low` [41,90) ·
+`opus-5-med` [90,100]. Effort tiers rank
+`low < medium < high < xhigh < max`.
+
+## Fixed model assignments (not scored)
+
+- **Orchestrator (you):** Opus 4.8 high — or Opus 5.5 / Fable 5 at medium+, the
+  two models that clear the guard from medium up. Never downgrade yourself.
+- **Review** — Superpowers path: Sonnet 4.6, medium. OpenSpec/self path:
+  Opus 5.5, low (`orc-reviewer-opus-5-low`; medium until v2.1.0).
+- **Verify:** Opus 5.5, medium (`orc-verifier-opus-5-med`).
+- **Analyst:** Opus 5.5, high. **Planner:** Opus 5.5, medium. **Test author:**
+  Opus 5.5, medium. **Combiner:** Opus 5.5, high. **Ultra advisor/judge:** Opus 5.5,
+  xhigh — every core fixed role is pinned to Opus 5.5 as of v0.34.0.
+- **Merge-conflict resolver:** Opus 4.8, medium.
+
+Note: every band in the table above is a real dispatch target (sonnet-5 through
+opus-5). If a model tier is unavailable in the environment, fall back UP to the
+next capable tier (never silently substitute a different family/effort).
+
+## Dispatch log (lives in the checkpoint)
+
+Every spawn records: `{task_id, computed_score, override_score|null,
+override_reason|null, model, effort, spawned_at}`. Feeds the usage report at
+every stop.

package/templates/skills/orc/schemas/planning-output.md

@@ -50,6 +50,9 @@ open_questions: [object]   # [] or [{question, proposed_default, blocking: bool}
   declared_files: [string] # HARD GATE: actual files it will touch, incl. tests.
                            # File-level globs. No declaration → no wave assignment.
   depends_on: [string]     # forward = ordering; reverse = stale_review flagging
+  precedent: [string]      # OPTIONAL. Read-only files this task copies behaviour
+                           # or style from (max 5); the lane passes them to the
+                           # pack as `--refs=`. Absent = the map picks them.
 
   # grounding attestation (filled by the PLANNER — one entry per declared path)
   grounding: [object]      # {path, disposition: exists|new, evidence}

package/templates/skills/orc-aftermath/SKILL.md

@@ -144,22 +144,22 @@ preflight line and has no gate to honour.
 ## Calls
 
 **ONE catalogue, and it is not you:** `orc lane calls orc-aftermath --json` names every
-CLI call this lane makes, each with its exit-code contract, its cost, when to run
+SHARED CLI call this lane makes, each with its exit-code contract, its cost, when to run
 it, and what an EMPTY answer means. Never invent a spelling, never re-word an
 exit code, and never re-derive a state word — the CLI's state words are the only
 state words, and **an exit code is an ANSWER wherever that contract says so, not
-a failure**. A call the answer does not name is a call this lane does not make.
+a failure**. A call that neither it nor this skill names is a call this lane does not make.
 Exit ≠ 0 from the catalogue itself → say the CLI is unavailable and name the
-command you are about to run, out loud, before running it.
-
-## Rules — the anti-slop card (`../_shared/phases/rules.md`)
-
-`orc rules slice --lane orc-aftermath --json` is the ONLY assembler; never build
-the card here. It rides under the house rules and above the task —
-**house rules > your project's rules > ORC's own packs** — and its `line` prints
-VERBATIM at preflight. Returns gain `rules_applied[]`, `rules_conflicts[]` (a gap,
-never a silent choice) and `rules_overridden[]`.
-## Waiting mid-run (`/orc-wait`)
-
-Canonical: `../_shared/wait.md`. **`a lane that waits without a hand-back` has broken this contract.**
-Checkpoint **none** · safe point **read-only, seconds long**. Nothing here to checkpoint, so all three modes behave identically — say so rather than asking. Never begin a wait between a dispatch and its validated return, or before the smoke gate has reported.
+command you are about to run, out loud, before running it.
+
+## Rules — the anti-slop card (`../_shared/phases/rules.md`)
+
+`orc rules slice --lane orc-aftermath --json` is the ONLY assembler; never build
+the card here. It rides under the house rules and above the task —
+**house rules > your project's rules > ORC's own packs** — and its `line` prints
+VERBATIM at preflight. Returns gain `rules_applied[]`, `rules_conflicts[]` (a gap,
+never a silent choice) and `rules_overridden[]`.
+## Waiting mid-run (`/orc-wait`)
+
+Canonical: `../_shared/wait.md`. **`a lane that waits without a hand-back` has broken this contract.**
+Checkpoint **none** · safe point **read-only, seconds long**. Nothing here to checkpoint, so all three modes behave identically — say so rather than asking. Never begin a wait between a dispatch and its validated return, or before the smoke gate has reported.

package/templates/skills/orc-analyze/SKILL.md

@@ -210,8 +210,8 @@ exist). Menu rules, relatedness gate, combiner handling:
 
 ## Lane contract (`../_shared/lane-contract.md` — read it ONLY when a call exits ≠ 0)
 
-- **Calls:** `orc lane calls orc-analyze --json` names every call and its exit codes.
-  **An exit code is an ANSWER where it says so, not a failure.** Make no other call.
+- **Calls:** `orc lane calls orc-analyze --json` names shared calls and exit codes.
+  **An exit code is an ANSWER where it says so, not a failure.** Own calls are here; make no other.
 - **Config:** `orc lane config orc-analyze --json`. Obey `effective`, print every line
   in `announce[]` VERBATIM at preflight, and honour `stops[]` before wave 1.
   Never merge `.claude/orc.config.yaml` yourself (`../_shared/config-precedence.md`).

package/templates/skills/orc-analyze-mini/SKILL.md

@@ -92,8 +92,8 @@ Never build or spawn.
 
 ## Lane contract (`../_shared/lane-contract.md` — read it ONLY when a call exits ≠ 0)
 
-- **Calls:** `orc lane calls orc-analyze-mini --json` names every call and its exit codes.
-  **An exit code is an ANSWER where it says so, not a failure.** Make no other call.
+- **Calls:** `orc lane calls orc-analyze-mini --json` names shared calls and exit codes.
+  **An exit code is an ANSWER where it says so, not a failure.** Own calls are here; make no other.
 - **Config:** `orc lane config orc-analyze-mini --json`. Obey `effective`. Never merge
   `.claude/orc.config.yaml` yourself (`../_shared/config-precedence.md`). Nothing
   here is contested, gated or a stop: no preflight line, no gate to honour.