@hegemonart/get-design-done 1.25.0 → 1.27.0

package/reference/model-prices.md

@@ -1,25 +1,35 @@
-# Model Prices — Static Price Table
+# Model Prices — Router
 
-**Source of truth for `est_cost_usd` calculations** in the router (`skills/router/SKILL.md`) and budget-enforcer hook (`hooks/budget-enforcer.js`). Anthropic-only pricing. Update the table here when prices change — downstream calculators read this file.
+**Phase 26 D-08 router.** This file used to carry a single Anthropic-only price table. As of v1.26.0 it links to per-runtime sub-tables — one file per runtime under `reference/prices/`. Budget-enforcer + cost-aggregator load the sub-table for the active runtime (resolved via `scripts/lib/runtime-detect.cjs`) and tag every `events.jsonl` cost row with the runtime ID.
 
-## Pricing (USD per 1M tokens)
+For the model→tier mapping (which model name corresponds to opus/sonnet/haiku per runtime), see `reference/runtime-models.md`.
 
-| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
-|-------|------|--------------|---------------|----------------------|
-| claude-haiku-4-5 | haiku | 1.00 | 5.00 | 0.10 |
-| claude-sonnet-4-7 | sonnet | 3.00 | 15.00 | 0.30 |
-| claude-opus-4-7 | opus | 15.00 | 75.00 | 1.50 |
+## Per-runtime sub-tables
+
+| Runtime | Path | Status |
+|---------|------|--------|
+| Claude Code | [`reference/prices/claude.md`](./prices/claude.md) | canonical (v1.26.0) |
+| OpenAI Codex CLI | [`reference/prices/codex.md`](./prices/codex.md) | seed (v1.26.0; provenance `<TODO>`) |
+| Google Gemini CLI | [`reference/prices/gemini.md`](./prices/gemini.md) | seed (v1.26.0; provenance `<TODO>`) |
+| Alibaba Qwen CLI | [`reference/prices/qwen.md`](./prices/qwen.md) | seed (v1.26.0; provenance `<TODO>`) |
+| Kilo Code | [`reference/prices/kilo.md`](./prices/kilo.md) | stub |
+| GitHub Copilot CLI | [`reference/prices/copilot.md`](./prices/copilot.md) | stub |
+| Cursor | [`reference/prices/cursor.md`](./prices/cursor.md) | stub |
+| Windsurf | [`reference/prices/windsurf.md`](./prices/windsurf.md) | stub |
+| Antigravity | [`reference/prices/antigravity.md`](./prices/antigravity.md) | stub |
+| Augment Code | [`reference/prices/augment.md`](./prices/augment.md) | stub |
+| Trae | [`reference/prices/trae.md`](./prices/trae.md) | stub |
+| CodeBuddy | [`reference/prices/codebuddy.md`](./prices/codebuddy.md) | stub |
+| Cline | [`reference/prices/cline.md`](./prices/cline.md) | stub |
+| OpenCode | [`reference/prices/opencode.md`](./prices/opencode.md) | stub |
 
-## size_budget → conservative token ranges
+**Sub-table format:** every file under `reference/prices/` carries the same canonical header row:
 
-Agent frontmatter carries `size_budget: S|M|L|XL`. The router uses these conservative token ranges to compute a pre-spawn `est_cost_usd` without a live model call:
+```
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+```
 
-| size_budget | input_tokens (conservative max) | output_tokens (conservative max) |
-|-------------|----------------------------------|-----------------------------------|
-| S | 4000 | 1000 |
-| M | 10000 | 2500 |
-| L | 25000 | 6000 |
-| XL | 60000 | 15000 |
+Extra columns may be appended at the right edge by runtime adapter authors without breaking the parser (forward-compatible).
 
 ## Estimator formula
 
@@ -29,9 +39,20 @@ est_cost_usd =
   + (output_tokens / 1_000_000) * output_per_1m
 ```
 
-When `cache_hit: true` (see D-08), the hook re-runs the formula with `cached_input_per_1m` in place of `input_per_1m` for the input portion.
+When `cache_hit: true`, the formula re-runs with `cached_input_per_1m` in place of `input_per_1m` for the input portion. See `skills/router/SKILL.md` (D-08) for the cache-hit semantics.
+
+## Fallback chain (D-08)
+
+When a cost lookup misses (model not present in the runtime's sub-table, or runtime sub-table is a stub), `scripts/lib/budget-enforcer.cjs` falls back to `reference/prices/claude.md` and emits a `cost_lookup_fallback` event. This keeps the pipeline running on stub runtimes while authority-watcher (Phase 13.2) flags drift for follow-up.
+
+If `claude.md` ALSO misses the model, the spawn proceeds with `cost_usd: null` and a `cost_lookup_failed` event — the existing fail-open contract from Phase 20-13.
+
+## Transitional fallback (v1.25 and earlier)
+
+For v1.25.x and earlier the single Anthropic price table lived inline in this file. That table is preserved at `reference/prices/claude.md` byte-for-byte (as of the v1.26.0 split, modulo the surrounding prose). Hooks/skills that pinned to specific row strings should rebase those references to the new path.
 
 ## Update protocol
 
-1. Pricing change: update the table above; commit as `chore(reference): update Anthropic pricing YYYY-MM-DD`.
-2. size_budget revision: requires a Phase 11 reflector proposal under `[FRONTMATTER]` scope; do not hand-edit agent ranges.
+1. Pricing change for a single runtime: edit only that runtime's file in `reference/prices/`. Commit as `chore(reference/prices/<runtime>): update <runtime> pricing YYYY-MM-DD`.
+2. New runtime added to the 14-runtime map (`scripts/lib/install/runtimes.cjs` + `reference/runtime-models.md`): create `reference/prices/<runtime>.md`, add a row to the table above, and add a `reference/registry.json` entry under `type: "data"`.
+3. size_budget revisions: requires a Phase 11 reflector proposal under `[FRONTMATTER]` scope. Token ranges are runtime-neutral and live in `reference/prices/claude.md` as the canonical reference.

package/reference/peer-cli-capabilities.md

@@ -0,0 +1,151 @@
+# Peer-CLI Capabilities
+
+Last verified: 2026-04-29
+
+Authoritative capability matrix for the peer-CLI delegation layer (Phase 27).
+The registry at `scripts/lib/peer-cli/registry.cjs` reads this map (encoded as
+data in the `.cjs` source — this doc is the human-readable mirror) to decide
+which peer-CLI claims which agent role and which protocol to speak.
+
+If you change this matrix, you MUST also change `CAPABILITY_MATRIX` in
+`scripts/lib/peer-cli/registry.cjs` and re-run `tests/peer-cli-registry.test.cjs`.
+The two are version-locked by Phase 27 D-05.
+
+## Capability matrix
+
+Each peer claims a fixed set of agent roles. The registry refuses to dispatch
+a role to a peer that does not claim it — this prevents accidental
+mis-delegations like "let's try `design-auditor` against Qwen" that produce
+garbage output (Phase 27 CONTEXT D-05).
+
+| Peer | Protocol | Claimed roles | Slash commands (within peer CLI) |
+| --- | --- | --- | --- |
+| `codex` | ASP | `execute` | `/exec` |
+| `copilot` | ACP | `review`, `research` | `/review`, `/search` |
+| `cursor` | ACP | `debug`, `plan` | `/debug`, `/plan` |
+| `gemini` | ACP | `research`, `exploration` | `/search`, `/explore` |
+| `qwen` | ACP | `write` | `/write` |
+
+Slash-command translation lives in each per-peer adapter
+(`scripts/lib/peer-cli/adapters/<peer>.cjs`, landed by Plan 27-04). The
+registry never invokes slash commands directly — it routes the role and
+delegates prompt-prefixing + slash translation to the adapter.
+
+## Tie-breaking when two peers claim the same role
+
+`research` is claimed by both `gemini` and `copilot`. When the registry's
+`findPeerFor('research', tier)` runs, it walks peers in alphabetical order
+of peer ID — `codex` < `copilot` < `cursor` < `gemini` < `qwen` — and
+returns the first one that passes the health probe. So `copilot` wins
+`research` over `gemini` when both are installed and allowlisted.
+
+Users can override this by removing one of the two from
+`.design/config.json#peer_cli.enabled_peers` — the unlisted peer is treated
+as absent regardless of installation status (Phase 27 D-11 opt-in gating).
+
+## Opt-in gating (D-11)
+
+A peer is dispatched to ONLY when:
+
+1. The peer ID appears in `.design/config.json#peer_cli.enabled_peers`
+   (an array of allowlisted peer IDs). Default: `[]` — empty, nothing
+   dispatches.
+2. The peer's adapter module loads at `scripts/lib/peer-cli/adapters/<peer>.cjs`.
+3. The adapter's `peerBinary()` resolver returns a path that exists on disk.
+
+Failure on any of the three → registry returns `null` and the caller
+(session-runner, Plan 27-06) falls back to the local Anthropic SDK call.
+This is the **transparent-fallback** contract from Phase 27 D-07: a missing
+or broken peer must never break the cycle.
+
+## Per-peer notes
+
+### `codex` (ASP)
+
+OpenAI Codex CLI invoked as `codex app-server`. Speaks the App Server
+Protocol — thread-oriented, supports resume across calls (currently unused;
+v1.27 always starts fresh threads).
+
+- **Provenance:** runtime entry in `scripts/lib/install/runtimes.cjs` (`id: 'codex'`).
+- **Example invocation:** registry routes `(role='execute', tier='opus', text='apply this diff')`
+  through `adapters/codex.cjs`, which in turn drives `asp-client.cjs`.
+- **Known limitations:** Codex' app-server cold start is ~1-2s on macOS;
+  the broker (Plan 27-03) keeps the session warm across cycles to amortize.
+
+### `gemini` (ACP)
+
+Google Gemini CLI invoked in ACP mode. One-shot prompt per call from the
+host's perspective, but the adapter may multiplex many calls onto a single
+broker session.
+
+- **Provenance:** runtime entry in `scripts/lib/install/runtimes.cjs` (`id: 'gemini'`).
+- **Example invocation:** registry routes `(role='research', tier='sonnet', text='find prior art for X')`
+  through `adapters/gemini.cjs`.
+- **Known limitations:** rate-limit headers vary by Gemini auth tier. The
+  adapter surfaces a 429 as a peer-error → registry returns null → caller
+  falls back. No retry logic at the registry level.
+
+### `cursor` (ACP)
+
+Cursor's `cursor` CLI in ACP mode. Strong on `debug` and `plan` roles
+because Cursor's editor-side context tracking transfers well to those
+workflows.
+
+- **Provenance:** runtime entry in `scripts/lib/install/runtimes.cjs` (`id: 'cursor'`).
+- **Known limitations:** Cursor' ACP mode requires an active Cursor login
+  on the host; an unauthenticated session surfaces as a connect-time error.
+
+### `copilot` (ACP)
+
+GitHub Copilot CLI in ACP mode. Claims `review` and `research`. Tends to
+win `research` against `gemini` on alphabetical tie-break — users who
+prefer Gemini for research should remove `copilot` from `enabled_peers`.
+
+- **Provenance:** runtime entry in `scripts/lib/install/runtimes.cjs` (`id: 'copilot'`).
+- **Known limitations:** Copilot's `review` role expects a diff in the
+  prompt; the adapter handles the framing.
+
+### `qwen` (ACP)
+
+Alibaba Qwen Code CLI in ACP mode. Claims `write` only. Useful when the
+session-runner wants a long-form code-generation pass and the host runtime
+is on a low tier.
+
+- **Provenance:** runtime entry in `scripts/lib/install/runtimes.cjs` (`id: 'qwen'`).
+- **Known limitations:** Qwen's ACP implementation is the newest of the
+  five; expect occasional protocol-version drift. The adapter pins
+  `protocolVersion: '2025-06-18'` per `acp-client.cjs` defaults.
+
+## Adding a new peer
+
+To add a new peer-CLI to the matrix:
+
+1. Run the guided ladder in `skills/peer-cli-add/SKILL.md` (lands with
+   Plan 27-10). It walks the protocol-fit check, the role-claim audit, the
+   adapter scaffold, and the test coverage required.
+2. Append the peer to `CAPABILITY_MATRIX` in `scripts/lib/peer-cli/registry.cjs`
+   AND to the table at the top of this file. The two MUST stay in sync —
+   the test suite (`tests/peer-cli-registry.test.cjs`) asserts the matrix
+   shape.
+3. Add a `<peer>.cjs` adapter under `scripts/lib/peer-cli/adapters/`.
+4. Add a `peerBinary?: string` field on the corresponding entry in
+   `scripts/lib/install/runtimes.cjs` (Plan 27-11 introduces the field;
+   new peers added after that plan ships must include it).
+5. Update `tests/peer-cli-registry.test.cjs` and the phase-20 baseline locks.
+
+To temporarily disable a peer without removing the adapter, drop the peer
+ID from `.design/config.json#peer_cli.enabled_peers`. To permanently remove
+a peer, reverse the steps above plus update `tests/phase-27-baseline.test.cjs`.
+
+## Cross-references
+
+- `scripts/lib/peer-cli/registry.cjs` — central dispatch, single source of
+  truth for the capability matrix as code.
+- `scripts/lib/peer-cli/adapters/*.cjs` — per-peer thin adapters
+  (Plan 27-04).
+- `scripts/lib/peer-cli/{acp,asp}-client.cjs` — protocol clients
+  (Plans 27-01 / 27-02).
+- `scripts/lib/peer-cli/broker-lifecycle.cjs` — long-lived session per
+  `(peer, workspace)` (Plan 27-03).
+- Phase 27 CONTEXT.md — decision log including D-05 (this matrix),
+  D-07 (transparent fallback), D-11 (opt-in gating).

package/reference/peer-protocols.md

@@ -0,0 +1,266 @@
+# Peer-CLI Protocols — ACP + ASP Cheat Sheet
+
+**Phase 27 (v1.27.0).** This file is the protocol-level reference for gdd's peer-CLI delegation layer. If you're authoring a new peer adapter or debugging a protocol-level issue, start here.
+
+For ops-level guidance (when delegation fires, how to enable/disable, fallback diagnostics), see `docs/PEER-DELEGATION.md`.
+
+Protocol shapes are adapted from [`greenpolo/cc-multi-cli`](https://github.com/greenpolo/cc-multi-cli) under Apache 2.0 — see `NOTICE` for full attribution.
+
+---
+
+## Two protocols, two transports
+
+| Protocol | Used by | Transport | Lifecycle |
+|----------|---------|-----------|-----------|
+| **ACP** (Agent Client Protocol) | Gemini, Cursor, Copilot, Qwen | Line-delimited JSON-RPC over stdio | Per-prompt request/response |
+| **ASP** (App Server Protocol) | Codex | Line-delimited JSON-RPC over stdio | Thread-oriented, multi-turn |
+
+Both protocols use the same line-delimited JSON-RPC framing (one JSON message per `\n`-terminated line on stdin/stdout). Both can be wrapped by gdd's broker (`scripts/lib/peer-cli/broker-lifecycle.cjs`) for long-lived sessions per `(peer, workspace)`.
+
+Line-buffer overflow guard: 16 MiB per line (both clients reject lines longer than this with a structured error).
+
+---
+
+## ACP — Agent Client Protocol
+
+### Initialize handshake
+
+Client → server, first message after spawn:
+
+```json
+{
+  "id": 1,
+  "jsonrpc": "2.0",
+  "method": "initialize",
+  "params": {
+    "protocolVersion": "2025-06-18",
+    "clientCapabilities": {}
+  }
+}
+```
+
+Server → client (reply correlated by `id`):
+
+```json
+{
+  "id": 1,
+  "jsonrpc": "2.0",
+  "result": {
+    "protocolVersion": "2025-06-18",
+    "serverCapabilities": { "...": "..." }
+  }
+}
+```
+
+If `serverCapabilities.protocolVersion` does not match the client's, the client logs a `protocol_mismatch` event and aborts the session.
+
+### Prompt method
+
+Client → server:
+
+```json
+{
+  "id": 2,
+  "jsonrpc": "2.0",
+  "method": "prompt",
+  "params": {
+    "text": "Research best React state libs"
+  }
+}
+```
+
+Server → client streams notifications (no `id`, no `result`) until the final `result` for the prompt's `id`:
+
+```json
+{ "jsonrpc": "2.0", "method": "agent_message_chunk", "params": { "text": "..." } }
+{ "jsonrpc": "2.0", "method": "tool_call",          "params": { "tool": "...", "input": "..." } }
+{ "jsonrpc": "2.0", "method": "file_change",        "params": { "path": "...", "diff": "..." } }
+{ "id": 2, "jsonrpc": "2.0", "result": { "content": "...", "finish_reason": "stop", "usage": { "input_tokens": 1234, "output_tokens": 5678 } } }
+```
+
+Notifications are surfaced via the `onNotification` callback the gdd caller passes:
+
+```js
+const result = await acpClient.prompt('Research ...', {
+  onNotification: (n) => console.log(n.method, n.params),
+});
+```
+
+### Per-peer ACP entry points
+
+Each peer documents its own way to enter ACP mode:
+
+- **Gemini**: `gemini acp` (subcommand).
+- **Cursor**: `cursor-agent acp` (subcommand on the CLI binary, not the IDE).
+- **Copilot**: `copilot --acp` (flag).
+- **Qwen**: `qwen acp` (subcommand).
+
+Verify via the `peer-cli-add` skill's verification ladder (Step 1) before adding a new peer.
+
+---
+
+## ASP — App Server Protocol (Codex)
+
+### Service identification
+
+Client → server, declared during initial communication:
+
+- `service_name = "gdd_peer_delegation"` (the canonical service identifier gdd uses)
+- `experimentalRawEvents = false` (we don't want raw model-token events; just structured turn output)
+
+These fields appear in both `threadStart` params and (where relevant) in handshake metadata.
+
+### Thread lifecycle
+
+ASP is thread-oriented. Each conversation has a `threadId`; turns happen within a thread.
+
+#### threadStart
+
+Client → server:
+
+```json
+{
+  "id": 1,
+  "jsonrpc": "2.0",
+  "method": "threadStart",
+  "params": {
+    "service_name": "gdd_peer_delegation",
+    "experimentalRawEvents": false
+  }
+}
+```
+
+Server → client:
+
+```json
+{
+  "id": 1,
+  "jsonrpc": "2.0",
+  "result": {
+    "threadId": "thread-abc123"
+  }
+}
+```
+
+#### threadResume
+
+Useful for cross-cycle conversation continuity (out of scope for v1.27.0 — gdd always creates fresh threads per delegated call — but the API surface exists):
+
+```json
+{
+  "id": 2,
+  "jsonrpc": "2.0",
+  "method": "threadResume",
+  "params": { "threadId": "thread-abc123" }
+}
+```
+
+Server replies with the thread's current state (turn history, last-known result).
+
+#### turn
+
+Client → server:
+
+```json
+{
+  "id": 3,
+  "jsonrpc": "2.0",
+  "method": "turn",
+  "params": {
+    "threadId": "thread-abc123",
+    "text": "Execute the build command"
+  }
+}
+```
+
+Server streams turn-progress notifications, ends with a structured result:
+
+**Completion path:**
+
+```json
+{
+  "id": 3,
+  "jsonrpc": "2.0",
+  "result": {
+    "threadId": "thread-abc123",
+    "turnId": "turn-xyz",
+    "status": "complete",
+    "content": "...",
+    "usage": { "input_tokens": 1234, "output_tokens": 5678 }
+  }
+}
+```
+
+**Error path** (does NOT throw on the client side — resolves with the error structure):
+
+```json
+{
+  "id": 3,
+  "jsonrpc": "2.0",
+  "result": {
+    "threadId": "thread-abc123",
+    "turnId": "turn-xyz",
+    "status": "error",
+    "error": {
+      "code": "rate_limit",
+      "message": "Rate limit exceeded for thread-abc123"
+    }
+  }
+}
+```
+
+The caller decides retry vs fallback per session-runner contract — gdd's session-runner falls back to local Anthropic on `status: "error"` per D-07.
+
+### Codex ASP entry point
+
+`codex app-server` (subcommand on the Codex CLI binary).
+
+---
+
+## Common framing rules (both protocols)
+
+### Line-delimited JSON-RPC
+
+- Each message is a single JSON object on stdin/stdout, terminated by `\n`.
+- Multiple messages may arrive in one chunk → client buffers until `\n`.
+- One message may split across chunks → client buffers until `\n`.
+- Lines longer than **16 MiB** are rejected with a structured error (the line buffer overflows, the client tears down and rejects all pending promises).
+
+### Request/response correlation
+
+- Requests carry `id` (monotonic integer per session).
+- Responses carry the same `id` in `result` or `error`.
+- Notifications have no `id` and no `result` — they are routed to the active request's `onNotification` callback (each protocol allows only one "active" request at a time per session — half-duplex).
+
+### Process lifecycle
+
+- The peer process is spawned via `scripts/lib/peer-cli/spawn-cmd.cjs` (handles Windows `.cmd` EINVAL workaround per D-04).
+- The client connects directly OR through gdd's broker (`broker-lifecycle.cjs`) — both surfaces present the same `{initialize, prompt, close}` (ACP) or `{threadStart, threadResume, turn, close}` (ASP) API.
+- On process death mid-request, the client rejects the in-flight promise with a structured `{error_class: "process_exited"}` event for telemetry.
+
+---
+
+## Adding a new protocol
+
+gdd v1.27.0 ships only ACP and ASP. If a new peer speaks neither (e.g., a future REST-only or HTTP/2-streaming protocol), the path forward is:
+
+1. Document the gap in `.design/RESEARCH.md` for a future phase to scope a new protocol layer.
+2. Do **not** stretch ACP/ASP to fit — they're documented contracts, not generalist multiplexers.
+3. The `peer-cli-add` skill (Step 1's verification ladder) refuses to scaffold a peer that doesn't speak ACP or ASP — by design.
+
+A future phase may add a new `scripts/lib/peer-cli/<protocol>-client.cjs` mirror following the same shape (line-buffer + JSON-RPC framing if applicable, or whatever the new protocol natively uses).
+
+---
+
+## Cross-references
+
+- `scripts/lib/peer-cli/acp-client.cjs` — ACP client implementation.
+- `scripts/lib/peer-cli/asp-client.cjs` — ASP client implementation.
+- `scripts/lib/peer-cli/spawn-cmd.cjs` — Windows `.cmd` EINVAL workaround.
+- `scripts/lib/peer-cli/broker-lifecycle.cjs` — long-lived broker.
+- `scripts/lib/peer-cli/adapters/*.cjs` — per-peer thin wrappers.
+- `scripts/lib/peer-cli/registry.cjs` — central dispatch.
+- `tests/peer-cli-{acp,asp,spawn,registry,adapters}.test.cjs` — protocol-level tests.
+- `docs/PEER-DELEGATION.md` — ops guide.
+- `NOTICE` — Apache 2.0 attribution for cc-multi-cli.
+- `.planning/phases/27-peer-cli-delegation/CONTEXT.md` — decision lineage (D-01, D-02, D-03, D-04).

package/reference/prices/antigravity.md

@@ -0,0 +1,21 @@
+# Antigravity — Price Table (stub)
+
+**Runtime:** `antigravity` (Antigravity)
+**Phase 26 D-08 sub-table — STUB.** Placeholder so the price-table router (`reference/model-prices.md`) has a complete link list for all 14 runtimes. Runtime adapter authors fill this in with provenance citations in a later cycle.
+
+**Provenance:** `<TODO: confirm at https://antigravity.google/docs>` — pending.
+
+## Pricing (USD per 1M tokens)
+
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+|-------|------|--------------|---------------|----------------------|
+| _TBD_ | opus | <TODO> | <TODO> | <TODO> |
+| _TBD_ | sonnet | <TODO> | <TODO> | <TODO> |
+| _TBD_ | haiku | <TODO> | <TODO> | <TODO> |
+
+The budget-enforcer treats unparseable rows as missing and falls back to `reference/prices/claude.md` per the D-08 fallback chain.
+
+## Update protocol
+
+1. Confirm authoritative numbers at the runtime author's pricing docs and update; remove `<TODO>` tags.
+2. Add provenance citation matching the `reference/runtime-models.md` row for `id: "antigravity"`.

package/reference/prices/augment.md

@@ -0,0 +1,21 @@
+# Augment Code — Price Table (stub)
+
+**Runtime:** `augment` (Augment Code)
+**Phase 26 D-08 sub-table — STUB.** Placeholder so the price-table router (`reference/model-prices.md`) has a complete link list for all 14 runtimes. Runtime adapter authors fill this in with provenance citations in a later cycle.
+
+**Provenance:** `<TODO: confirm at https://docs.augmentcode.com>` — pending.
+
+## Pricing (USD per 1M tokens)
+
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+|-------|------|--------------|---------------|----------------------|
+| _TBD_ | opus | <TODO> | <TODO> | <TODO> |
+| _TBD_ | sonnet | <TODO> | <TODO> | <TODO> |
+| _TBD_ | haiku | <TODO> | <TODO> | <TODO> |
+
+The budget-enforcer treats unparseable rows as missing and falls back to `reference/prices/claude.md` per the D-08 fallback chain.
+
+## Update protocol
+
+1. Confirm authoritative numbers at the runtime author's pricing docs and update; remove `<TODO>` tags.
+2. Add provenance citation matching the `reference/runtime-models.md` row for `id: "augment"`.

package/reference/prices/claude.md

@@ -0,0 +1,42 @@
+# Anthropic — Claude Code Price Table
+
+**Runtime:** `claude` (Claude Code)
+**Phase 26 D-08 sub-table.** Authoritative pricing for the Anthropic models referenced in `reference/runtime-models.md` under `id: "claude"`. Read by `scripts/lib/budget-enforcer.cjs` (and indirectly by `hooks/budget-enforcer.ts`) to compute `est_cost_usd` per spawn.
+
+**Provenance:** https://docs.anthropic.com/en/docs/about-claude/pricing — retrieved 2026-04-29, cycle `2026-04-29-v1.26`.
+
+## Pricing (USD per 1M tokens)
+
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+|-------|------|--------------|---------------|----------------------|
+| claude-haiku-4-5 | haiku | 1.00 | 5.00 | 0.10 |
+| claude-sonnet-4-7 | sonnet | 3.00 | 15.00 | 0.30 |
+| claude-sonnet-4-6 | sonnet | 3.00 | 15.00 | 0.30 |
+| claude-opus-4-7 | opus | 15.00 | 75.00 | 1.50 |
+
+## size_budget → conservative token ranges
+
+Agent frontmatter carries `size_budget: S|M|L|XL`. The router uses these conservative token ranges to compute a pre-spawn `est_cost_usd` without a live model call:
+
+| size_budget | input_tokens (conservative max) | output_tokens (conservative max) |
+|-------------|----------------------------------|-----------------------------------|
+| S | 4000 | 1000 |
+| M | 10000 | 2500 |
+| L | 25000 | 6000 |
+| XL | 60000 | 15000 |
+
+## Estimator formula
+
+```
+est_cost_usd =
+  (input_tokens / 1_000_000) * input_per_1m
+  + (output_tokens / 1_000_000) * output_per_1m
+```
+
+When `cache_hit: true`, the formula re-runs with `cached_input_per_1m` in place of `input_per_1m` for the input portion.
+
+## Update protocol
+
+1. Pricing change: update the table above; commit as `chore(reference/prices): update Anthropic pricing YYYY-MM-DD`.
+2. New model name added to `reference/runtime-models.md` under `id: "claude"`: add a row here with the same model string in the `Model` column. Tier comes from the canonical `tier_to_model` mapping.
+3. size_budget revision: requires a Phase 11 reflector proposal under `[FRONTMATTER]` scope; do not hand-edit agent ranges.

package/reference/prices/cline.md

@@ -0,0 +1,23 @@
+# Cline — Price Table (stub)
+
+**Runtime:** `cline` (Cline)
+**Phase 26 D-08 sub-table — STUB.** Placeholder so the price-table router (`reference/model-prices.md`) has a complete link list for all 14 runtimes. Runtime adapter authors fill this in with provenance citations in a later cycle.
+
+**Provenance:** `<TODO: confirm at https://docs.cline.bot>` — pending.
+
+**Note:** Cline is a BYO-API-key runtime — actual pricing is determined by whichever provider key the user supplies. The `tier_to_model` row in `reference/runtime-models.md` chooses a default model per tier; this table prices that default. Users of other providers should override locally.
+
+## Pricing (USD per 1M tokens)
+
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+|-------|------|--------------|---------------|----------------------|
+| _TBD_ | opus | <TODO> | <TODO> | <TODO> |
+| _TBD_ | sonnet | <TODO> | <TODO> | <TODO> |
+| _TBD_ | haiku | <TODO> | <TODO> | <TODO> |
+
+The budget-enforcer treats unparseable rows as missing and falls back to `reference/prices/claude.md` per the D-08 fallback chain.
+
+## Update protocol
+
+1. Confirm authoritative numbers at the runtime author's pricing docs and update; remove `<TODO>` tags.
+2. Add provenance citation matching the `reference/runtime-models.md` row for `id: "cline"`.

package/reference/prices/codebuddy.md

@@ -0,0 +1,21 @@
+# CodeBuddy — Price Table (stub)
+
+**Runtime:** `codebuddy` (Tencent CodeBuddy)
+**Phase 26 D-08 sub-table — STUB.** Placeholder so the price-table router (`reference/model-prices.md`) has a complete link list for all 14 runtimes. Runtime adapter authors fill this in with provenance citations in a later cycle.
+
+**Provenance:** `<TODO: confirm at https://copilot.tencent.com>` — pending.
+
+## Pricing (USD per 1M tokens)
+
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+|-------|------|--------------|---------------|----------------------|
+| _TBD_ | opus | <TODO> | <TODO> | <TODO> |
+| _TBD_ | sonnet | <TODO> | <TODO> | <TODO> |
+| _TBD_ | haiku | <TODO> | <TODO> | <TODO> |
+
+The budget-enforcer treats unparseable rows as missing and falls back to `reference/prices/claude.md` per the D-08 fallback chain.
+
+## Update protocol
+
+1. Confirm authoritative numbers at the runtime author's pricing docs and update; remove `<TODO>` tags.
+2. Add provenance citation matching the `reference/runtime-models.md` row for `id: "codebuddy"`.

package/reference/prices/codex.md

@@ -0,0 +1,25 @@
+# OpenAI — Codex CLI Price Table
+
+**Runtime:** `codex` (OpenAI Codex CLI)
+**Phase 26 D-08 sub-table.** Pricing for the OpenAI Codex tier referenced in `reference/runtime-models.md` under `id: "codex"`.
+
+**Provenance:** `<TODO: confirm at https://openai.com/api/pricing/>` — retrieved 2026-04-29 (placeholder — v1.26.0 ships with seed numbers; runtime adapter authors confirm and PR before v1.27).
+
+**Status:** placeholder values are taken from public OpenAI tier-positioning at the time of v1.26.0 ship. The cost-aggregator will surface drift if measured spend per spawn diverges from these figures by more than 20% after 10+ cycles.
+
+## Pricing (USD per 1M tokens)
+
+| Model | Tier | input_per_1m | output_per_1m | cached_input_per_1m |
+|-------|------|--------------|---------------|----------------------|
+| gpt-5 | opus | 1.25 | 10.00 | 0.13 |
+| gpt-5-mini | sonnet | 0.25 | 2.00 | 0.03 |
+| gpt-5-nano | haiku | 0.05 | 0.40 | 0.01 |
+
+## Estimator formula
+
+Same shape as `reference/prices/claude.md`; see that file for the formula and `size_budget` ranges. Token ranges are runtime-neutral.
+
+## Update protocol
+
+1. Confirm authoritative numbers at https://openai.com/api/pricing/ and update the table; remove the `<TODO>` provenance tag.
+2. New model added to `reference/runtime-models.md` under `id: "codex"`: add a row here with the matching model string and tier.