@kontextmind/kxm 0.7.209 → 0.7.211

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (17) hide show
  1. package/.claude-plugin/marketplace.json +1 -1
  2. package/.kxm/roles/reviewer-cli.yaml +3 -1
  3. package/CHANGELOG.md +20 -0
  4. package/docs/contributing/assignment-runner.md +1 -1
  5. package/docs/contributing/harness-routing-internals.md +6 -4
  6. package/docs/reference/cli-reference.md +3 -1
  7. package/docs/reference/config-reference.md +8 -1
  8. package/docs/reference/harness-routing.md +1 -1
  9. package/docs/reference/workflow-catalog.md +49 -50
  10. package/package.json +1 -1
  11. package/plugins/kxm/.claude-plugin/plugin.json +1 -1
  12. package/plugins/kxm/dist/cli.js +16 -17
  13. package/plugins/kxm/dist/mcp-server.js +1 -1
  14. package/plugins/kxm/package.json +1 -1
  15. package/plugins/kxm/src/init-guide-setup.ts +16 -17
  16. package/plugins/kxm/src/mcp-server.ts +1 -1
  17. package/scripts/harness-run.mjs +1 -1
package/.claude-plugin/marketplace.json CHANGED
@@ -11,7 +11,7 @@
11
11
  "name": "kxm",
12
12
  "source": "./plugins/kxm",
13
13
  "description": "Durable workflows, peer agents, and kxm tui",
14
- "version": "0.7.209",
14
+ "version": "0.7.211",
15
15
  "category": "development",
16
16
  "tags": ["kxm", "multi-agent", "workflows", "mcp"]
17
17
  }
package/.kxm/roles/reviewer-cli.yaml CHANGED
@@ -4,7 +4,7 @@ purpose: reviewer-cli
4
4
  permission: read-only
5
5
  description: Independent CLI and docs critic.
6
6
  # codex-gpt-6-1-sol is the required cli critic. The Pi fallback is a different vendor.
7
- # codex-gpt-5-6-sol stays admitted as the last fallback.
7
+ # codex-gpt-5-6-sol stays the backup. codex-gpt-6-1-terra is the third fallback.
8
8
  roster:
9
9
  - route: codex-gpt-6-1-sol
10
10
  effort: low
@@ -12,3 +12,5 @@ roster:
12
12
  effort: low
13
13
  - route: codex-gpt-5-6-sol
14
14
  effort: low
15
+ - route: codex-gpt-6-1-terra
16
+ effort: low
package/CHANGELOG.md CHANGED
@@ -225,6 +225,26 @@ All notable user-facing changes are documented here. The project follows [Semant
225
225
 
226
226
  ### Changed
227
227
 
228
+ - **Workflow catalog prices were re-verified against OpenRouter on 2026-11-04.**
229
+ Candidate prices in [the workflow catalog](docs/reference/workflow-catalog.md)
230
+ match `https://openrouter.ai/api/v1/models` from that day.
231
+ `deepseek/deepseek-v4-pro-0813` is $0.115 input and $4.60 output per 1M
232
+ tokens, so it leaves the Tier 1 workhorse picks for `qwen/qwen3.7-plus` or
233
+ `google/gemini-3.8-flash`. The adversarial security slot names
234
+ `moonshotai/kimi-k3.1-pro` ($2 / $10), which also replaces `kimi-k3` on long
235
+ agentic work. It is slow and not interactive. `openai/gpt-6.1-terra`
236
+ ($1 / $5, then $2 / $7.50 above 272,000 prompt tokens) is the cost-sensitive
237
+ Tier 1/2 workhorse and the third `reviewer-cli` fallback. Primary stays
238
+ GPT-6.1 Sol and the backup stays GPT-5.6 Sol. The planner stays Fable, then
239
+ Sonnet 5.5. The writer roster is unchanged. `openai/gpt-6-luna` ($0.10 / $0.50)
240
+ replaces GPT-5.6 Luna in Tier 0. `anthropic/claude-haiku-5.5` ($0.10 / $0.50,
241
+ 1M context) replaces Haiku 4.5. `poolside/laguna-s-2.1` is gone; that slot
242
+ names `qwen/qwen3.8-flash`. `z-ai/glm-5.3-flash` is $0.0269 / $0.0996 and
243
+ the list price is volatile. `z-ai/glm-5.3` is $0.08 / $3.60.
244
+ `.kxm/prices.yaml` is dated 2026-11-04. The retired
245
+ `qwen/qwen3-coder-plus` row stays marked historical. The weekly refresh
246
+ still does not rewrite `zai-coding-cn/glm-5.3` or the flash coding-plan row.
247
+
228
248
  - **Workflow catalog prices were re-verified against OpenRouter on 2026-10-29.**
229
249
  Candidate context windows and list prices in
230
250
  [the workflow catalog](docs/reference/workflow-catalog.md) match
package/docs/contributing/assignment-runner.md CHANGED
@@ -37,7 +37,7 @@ admits each route for specific roles and permissions:
37
37
  | `writer` | `grok` / `grok-4.7`; relief: `pi` / `openrouter/qwen/qwen3.7-plus`; `agy` / `gemini-4-argon-high`, then `gemini-4-argon-medium`, then `gemini-3.8-flash-medium`, then `gemini-3.8-flash-high` | `xai`; `alibaba`; `google` | `edit` |
38
38
  | `planner` | `claude` / `fable`, then `claude` / `claude-sonnet-5.5`; fallback `pi` / `openrouter/qwen/qwen3.8-flash` | `anthropic`; `anthropic`; `alibaba` | `read-only` |
39
39
  | `reviewer-arch` | `claude` / `fable` | `anthropic` | `read-only` |
40
- | `reviewer-cli` | `codex` / `gpt-6.1-sol`; fallback `pi` / `openrouter/qwen/qwen3.8-flash`, then `codex` / `gpt-5.6-sol` | `openai`; `alibaba`; `openai` | `read-only` |
40
+ | `reviewer-cli` | `codex` / `gpt-6.1-sol`; fallback `pi` / `openrouter/qwen/qwen3.8-flash`, then `codex` / `gpt-5.6-sol`, then `codex` / `gpt-6.1-terra` | `openai`; `alibaba`; `openai`; `openai` | `read-only` |
41
41
 
42
42
  The policy itself enforces three rules. Critics are read-only. The two
43
43
  designated critics have different vendors. No admitted writer shares a vendor
package/docs/contributing/harness-routing-internals.md CHANGED
@@ -9,7 +9,7 @@ This page records how the KXM repository applies [harness routing](../reference/
9
9
 
10
10
  The files use one id shape per kind. A route id is `<harness>-<model-slug>[-<provider>]`, with `.` written as `-`. An agent id equals its role. A workflow id is `default`, `land`, or `<role>-only`. An agent step id equals that role.
11
11
 
12
- This checkout's admitted routes are `grok-grok-4-7`, `pi-qwen3-7-plus-openrouter`, `agy-gemini-4-argon-high`, `agy-gemini-4-argon-medium`, `agy-gemini-3-8-flash-medium`, `agy-gemini-3-8-flash-high`, `claude-fable`, `claude-claude-sonnet-5-5`, `codex-gpt-6-1-sol`, `codex-gpt-5-6-sol`, `pi-qwen3-8-flash-openrouter`, and `pi-glm-5-3-flash-openrouter`. `pi-qwen3-coder-plus-openrouter` stays on disk with `status: retired` because OpenRouter removed `qwen/qwen3-coder-plus` (catalog 404); it is not admitted and not on a roster, and `qwen-openrouter-pi` does not resolve. Agents are `planner`, `writer`, `reviewer-arch`, and `reviewer-cli`. `opus-claude` was removed because `opus` is not an admitted selector. `qwen-token-plan/*` and `zai-coding-cn/*` left the admitted list because those harnesses are not allowlisted; their `prices.yaml` rows stay. Other old ids resolve as aliases. Every roster entry names an effort. `reviewer-arch` dispatches `claude-fable`. #343's architecture critic ran on opus because the justfile recipe review-arch hardcoded that model; the role file on that commit listed only `fable-claude`. The helper now refuses `opus`.
12
+ This checkout's admitted routes are `grok-grok-4-7`, `pi-qwen3-7-plus-openrouter`, `agy-gemini-4-argon-high`, `agy-gemini-4-argon-medium`, `agy-gemini-3-8-flash-medium`, `agy-gemini-3-8-flash-high`, `claude-fable`, `claude-claude-sonnet-5-5`, `codex-gpt-6-1-sol`, `codex-gpt-5-6-sol`, `codex-gpt-6-1-terra`, `pi-qwen3-8-flash-openrouter`, and `pi-glm-5-3-flash-openrouter`. `pi-qwen3-coder-plus-openrouter` stays on disk with `status: retired` because OpenRouter removed `qwen/qwen3-coder-plus` (catalog 404); it is not admitted and not on a roster, and `qwen-openrouter-pi` does not resolve. Agents are `planner`, `writer`, `reviewer-arch`, and `reviewer-cli`. `opus-claude` was removed because `opus` is not an admitted selector. `qwen-token-plan/*` and `zai-coding-cn/*` left the admitted list because those harnesses are not allowlisted; their `prices.yaml` rows stay. Other old ids resolve as aliases. Every roster entry names an effort. `reviewer-arch` dispatches `claude-fable`. #343's architecture critic ran on opus because the justfile recipe review-arch hardcoded that model; the role file on that commit listed only `fable-claude`. The helper now refuses `opus`.
13
13
 
14
14
  ## This checkout's routes and roster
15
15
 
@@ -37,7 +37,7 @@ roster:
37
37
  effort: medium
38
38
  ```
39
39
 
40
- The writer role lists route ids, in order: `grok-grok-4-7`, then `pi-qwen3-7-plus-openrouter`, then `agy-gemini-4-argon-high`, `agy-gemini-4-argon-medium`, `agy-gemini-3-8-flash-medium`, and `agy-gemini-3-8-flash-high`. `grok-grok-4-7` is harness `grok`, model `grok-4.7`, vendor `xai`. An agent file does not declare `harness:`. Argon is writer #3 and #4. Flash medium is the safety net because Argon needs a paid Google AI Pro, Ultra, or Enterprise plan and `agy -p` exits when the pinned model is missing. Flash high stays on the roster so that admitted model file is still used. The planner roster is `claude-fable`, then `claude-claude-sonnet-5-5` (`anthropic/claude-sonnet-5.5`), then `pi-qwen3-8-flash-openrouter`. Fable stays the primary. The [October 2026 planner bake-off](../evals/planner-bakeoff-2026-10.md) is why Sonnet is the fallback. The CLI critic primary is `codex-gpt-6-1-sol`, then `pi-qwen3-8-flash-openrouter`, then `codex-gpt-5-6-sol`. The architecture critic primary stays `claude-fable`, then `pi-glm-5-3-flash-openrouter`. Argon can serve as a planner or architecture-review fallback; it is not on those rosters. GPT-6.1 Sol Pro is the same model with reasoning mode `pro`. The route id does not carry a reasoning mode, role effort has no `pro` value, and codex is limited to `reviewer-cli`, so that mode is not an architecture fallback.
40
+ The writer role lists route ids, in order: `grok-grok-4-7`, then `pi-qwen3-7-plus-openrouter`, then `agy-gemini-4-argon-high`, `agy-gemini-4-argon-medium`, `agy-gemini-3-8-flash-medium`, and `agy-gemini-3-8-flash-high`. `grok-grok-4-7` is harness `grok`, model `grok-4.7`, vendor `xai`. An agent file does not declare `harness:`. Argon is writer #3 and #4. Flash medium is the safety net because Argon needs a paid Google AI Pro, Ultra, or Enterprise plan and `agy -p` exits when the pinned model is missing. Flash high stays on the roster so that admitted model file is still used. The planner roster is `claude-fable`, then `claude-claude-sonnet-5-5` (`anthropic/claude-sonnet-5.5`), then `pi-qwen3-8-flash-openrouter`. Fable stays the primary. The [October 2026 planner bake-off](../evals/planner-bakeoff-2026-10.md) is why Sonnet is the fallback. The CLI critic primary is `codex-gpt-6-1-sol`, then `pi-qwen3-8-flash-openrouter`, then `codex-gpt-5-6-sol`, then `codex-gpt-6-1-terra`. The architecture critic primary stays `claude-fable`, then `pi-glm-5-3-flash-openrouter`. Argon can serve as a planner or architecture-review fallback; it is not on those rosters. GPT-6.1 Sol Pro is the same model with reasoning mode `pro`. The route id does not carry a reasoning mode, role effort has no `pro` value, and codex is limited to `reviewer-cli`, so that mode is not an architecture fallback.
41
41
 
42
42
  `node scripts/kxm.mjs routes list`:
43
43
 
@@ -50,6 +50,7 @@ admitted google/gemini-4-argon-high
50
50
  admitted google/gemini-4-argon-medium
51
51
  admitted openai/gpt-5.6-sol
52
52
  admitted openai/gpt-6.1-sol
53
+ admitted openai/gpt-6.1-terra
53
54
  admitted openrouter/qwen/qwen3.7-plus
54
55
  admitted openrouter/qwen/qwen3.8-flash
55
56
  admitted openrouter/z-ai/glm-5.3-flash
@@ -62,6 +63,7 @@ reviewer-arch pi-glm-5-3-flash-openrouter
62
63
  reviewer-cli codex-gpt-6-1-sol
63
64
  reviewer-cli pi-qwen3-8-flash-openrouter
64
65
  reviewer-cli codex-gpt-5-6-sol
66
+ reviewer-cli codex-gpt-6-1-terra
65
67
  writer grok-grok-4-7
66
68
  writer pi-qwen3-7-plus-openrouter
67
69
  writer agy-gemini-4-argon-high
@@ -142,7 +144,7 @@ Pi's own `xai` provider reported `ready` (OAuth) on the capture machine. When th
142
144
  | Billing | ChatGPT subscription. List price $2.00 input, $10.00 output, $0.10 cached input, $2.50 cache write through 272,000 prompt tokens, then $4.00 / $15.00 | Same per-token list price. OpenRouter's `openai/gpt-6.1-sol-pro` row is reasoning mode `pro`, not a second model |
143
145
  | Context | 1,050,000 (922,000 max input, 128,000 max output). Reasoning is mandatory | 1,050,000 |
144
146
 
145
- `codex-gpt-5-6-sol` stays on the `reviewer-cli` roster after the Qwen fallback. Its selector `openai/gpt-5.6-sol` stays admitted. OpenRouter lists that model at $4.00 input, $20.00 output, and $0.40 cached input. `pi auth check --provider openai-codex` reported `ready` on the capture machine. That readiness is not permission to bill OpenAI through Pi.
147
+ `codex-gpt-5-6-sol` stays on the `reviewer-cli` roster after the Qwen fallback. Its selector `openai/gpt-5.6-sol` stays admitted. OpenRouter lists that model at $4.00 input, $20.00 output, and $0.40 cached input. `codex-gpt-6-1-terra` is the third fallback after that backup. OpenRouter on 2026-11-04 lists `openai/gpt-6.1-terra` at $1.00 input and $5.00 output through 272,000 prompt tokens, then $2.00 and $7.50. `openrouter/z-ai/glm-5.3-flash` was re-verified the same day at $0.0269 input, $0.0996 output, and $0.0269 cache read. That list price is volatile. `pi auth check --provider openai-codex` reported `ready` on the capture machine. That readiness is not permission to bill OpenAI through Pi.
146
148
 
147
149
  ### Claude Fable
148
150
 
@@ -182,7 +184,7 @@ The code differs from that decision: `.kxm/routes.yaml` admits `google/gemini-3.
182
184
  |---|---|---|---|
183
185
  | Qwen3.8 Flash | `qwen-token-plan/qwen3.8-flash`: removed from the admitted list on 2026-09-27; `qwen-token-plan` is not allowlisted | `openrouter/qwen/qwen3.8-flash`: admitted, route `pi-qwen3-8-flash-openrouter`, read-only fallback for planner and reviewer-cli | `prices.yaml` keeps both rows. |
184
186
  | Qwen3.7 Plus | none | `openrouter/qwen/qwen3.7-plus`: route `pi-qwen3-7-plus-openrouter`, writer #2 | The only admitted Pi writer: exact model, `edit` permission. OpenRouter list on 2026-10-17 is $0.32 input and $1.28 output per 1M tokens, 1,000,000 context, tools supported. |
185
- | GLM 5.3 and 5.3 Flash | `zai-coding-cn/glm-5.3` and `…/glm-5.3-flash`: removed from the admitted list on 2026-09-27; `zai-coding-cn` is not allowlisted. The flash coding-plan row keeps $0.15 / $0.50 (cache read $0.03). `zai-coding-cn/glm-5.3` output is $1.40 | `openrouter/z-ai/glm-5.3-flash`: admitted, route `pi-glm-5-3-flash-openrouter`, read-only fallback for reviewer-arch. Its own tier, verified 2026-10-21: $0.15 input, $0.50 output, $0.03 cache read. This list price is volatile and has flipped with $0.047076 / $0.12 / $0.006 for several days; trust the weekly refresh over the static row | Z.ai has no native harness. The two flash rows do not share a YAML anchor. |
187
+ | GLM 5.3 and 5.3 Flash | `zai-coding-cn/glm-5.3` and `…/glm-5.3-flash`: removed from the admitted list on 2026-09-27; `zai-coding-cn` is not allowlisted. The flash coding-plan row keeps $0.15 / $0.50 (cache read $0.03). `zai-coding-cn/glm-5.3` follows the 2026-11-04 public list price, $0.08 input and $3.60 output (cache read $0.076), and the weekly refresh still does not rewrite it | `openrouter/z-ai/glm-5.3-flash`: admitted, route `pi-glm-5-3-flash-openrouter`, read-only fallback for reviewer-arch. Its own tier, verified 2026-11-04: $0.0269 input, $0.0996 output, $0.0269 cache read. This list price is volatile and has also listed $0.047076 / $0.12 / $0.006 and $0.15 / $0.50; trust the weekly refresh over the static row | Z.ai has no native harness. The two flash rows do not share a YAML anchor. |
186
188
  | DeepSeek V4.1 Flash | `qwen-token-plan/deepseek-v4.1-flash`: removed from the admitted list on 2026-09-27 | `deepseek/deepseek-v4.1-flash` in the OpenRouter feed | Bills a DeepSeek model through Alibaba's plan. `deepseek` is a native-vendor Pi brake, so this is not a route. |
187
189
 
188
190
  The developer runner is stricter. Its only Pi writer is `openrouter/qwen/qwen3.7-plus`, and it does not allowlist `qwen-token-plan` or `zai-coding-cn`. Today OpenRouter is the right answer here for Qwen3.7 Plus, Qwen3.8 Flash and GLM 5.3 Flash. `qwen/qwen3-coder-plus` is retired.
package/docs/reference/cli-reference.md CHANGED
@@ -1190,6 +1190,7 @@ admitted google/gemini-4-argon-high
1190
1190
  admitted google/gemini-4-argon-medium
1191
1191
  admitted openai/gpt-5.6-sol
1192
1192
  admitted openai/gpt-6.1-sol
1193
+ admitted openai/gpt-6.1-terra
1193
1194
  admitted openrouter/qwen/qwen3.7-plus
1194
1195
  admitted openrouter/qwen/qwen3.8-flash
1195
1196
  admitted openrouter/z-ai/glm-5.3-flash
@@ -1202,6 +1203,7 @@ reviewer-arch pi-glm-5-3-flash-openrouter
1202
1203
  reviewer-cli codex-gpt-6-1-sol
1203
1204
  reviewer-cli pi-qwen3-8-flash-openrouter
1204
1205
  reviewer-cli codex-gpt-5-6-sol
1206
+ reviewer-cli codex-gpt-6-1-terra
1205
1207
  writer grok-grok-4-7
1206
1208
  writer pi-qwen3-7-plus-openrouter
1207
1209
  writer agy-gemini-4-argon-high
@@ -1215,7 +1217,7 @@ kxm routes list --json
1215
1217
  ```
1216
1218
 
1217
1219
  ```text
1218
- {"schema":"kxm.cli-result.v1","ok":true,"command":"routes list","policy":{"schema":"kxm.routes.v2","updatedAt":"2026-10-20T22:40:00.000Z","admitted":["anthropic/claude-sonnet-5.5","anthropic/fable","google/gemini-3.8-flash-high","google/gemini-3.8-flash-medium","google/gemini-4-argon-high","google/gemini-4-argon-medium","openai/gpt-5.6-sol","openai/gpt-6.1-sol","openrouter/qwen/qwen3.7-plus","openrouter/qwen/qwen3.8-flash","openrouter/z-ai/glm-5.3-flash","xai/grok-4.7"],"disabled":[]},"membership":["planner claude-fable","planner claude-claude-sonnet-5-5","planner pi-qwen3-8-flash-openrouter","reviewer-arch claude-fable","reviewer-arch pi-glm-5-3-flash-openrouter","reviewer-cli codex-gpt-6-1-sol","reviewer-cli pi-qwen3-8-flash-openrouter","reviewer-cli codex-gpt-5-6-sol","writer grok-grok-4-7","writer pi-qwen3-7-plus-openrouter","writer agy-gemini-4-argon-high","writer agy-gemini-4-argon-medium","writer agy-gemini-3-8-flash-medium","writer agy-gemini-3-8-flash-high"]}
1220
+ {"schema":"kxm.cli-result.v1","ok":true,"command":"routes list","policy":{"schema":"kxm.routes.v2","updatedAt":"2026-11-04T00:00:00.000Z","admitted":["anthropic/claude-sonnet-5.5","anthropic/fable","google/gemini-3.8-flash-high","google/gemini-3.8-flash-medium","google/gemini-4-argon-high","google/gemini-4-argon-medium","openai/gpt-5.6-sol","openai/gpt-6.1-sol","openai/gpt-6.1-terra","openrouter/qwen/qwen3.7-plus","openrouter/qwen/qwen3.8-flash","openrouter/z-ai/glm-5.3-flash","xai/grok-4.7"],"disabled":[]},"membership":["planner claude-fable","planner claude-claude-sonnet-5-5","planner pi-qwen3-8-flash-openrouter","reviewer-arch claude-fable","reviewer-arch pi-glm-5-3-flash-openrouter","reviewer-cli codex-gpt-6-1-sol","reviewer-cli pi-qwen3-8-flash-openrouter","reviewer-cli codex-gpt-5-6-sol","reviewer-cli codex-gpt-6-1-terra","writer grok-grok-4-7","writer pi-qwen3-7-plus-openrouter","writer agy-gemini-4-argon-high","writer agy-gemini-4-argon-medium","writer agy-gemini-3-8-flash-medium","writer agy-gemini-3-8-flash-high"]}
1219
1221
  ```
1220
1222
 
1221
1223
  ### `kxm routes count`
package/docs/reference/config-reference.md CHANGED
@@ -1105,11 +1105,18 @@ routes:
1105
1105
  roles: [reviewer-cli]
1106
1106
  permissions: [read-only]
1107
1107
  status: admitted
1108
+ codex-gpt-6-1-terra:
1109
+ harness: codex
1110
+ model: gpt-6.1-terra
1111
+ vendor: openai
1112
+ roles: [reviewer-cli]
1113
+ permissions: [read-only]
1114
+ status: admitted
1108
1115
  lineup: # routes admitted for each role, preference order
1109
1116
  writer: [grok-grok-4-7, pi-qwen3-7-plus-openrouter, agy-gemini-4-argon-high, agy-gemini-4-argon-medium, agy-gemini-3-8-flash-medium, agy-gemini-3-8-flash-high]
1110
1117
  planner: [claude-fable, claude-claude-sonnet-5-5, pi-qwen3-8-flash-openrouter]
1111
1118
  reviewer-arch: [claude-fable, pi-glm-5-3-flash-openrouter]
1112
- reviewer-cli: [codex-gpt-6-1-sol, pi-qwen3-8-flash-openrouter, codex-gpt-5-6-sol]
1119
+ reviewer-cli: [codex-gpt-6-1-sol, pi-qwen3-8-flash-openrouter, codex-gpt-5-6-sol, codex-gpt-6-1-terra]
1113
1120
  required_critics:
1114
1121
  review-arch: claude-fable
1115
1122
  review-cli: codex-gpt-6-1-sol
package/docs/reference/harness-routing.md CHANGED
@@ -378,7 +378,7 @@ pi auth check --provider xai --json --no-refresh
378
378
  | Recorded `costBasis` | `unmetered`, `priceRef: subscription:codex` | None: refused before dispatch |
379
379
  | Mode | One-shot `codex exec --json -`. Reasoning is mandatory (`low`, `medium`, `high`, `xhigh`, `max`) | Refused |
380
380
 
381
- `codex-gpt-6-1-sol` is the `reviewer-cli` primary at effort `low`. `pi-qwen3-8-flash-openrouter` is the other-vendor fallback. `codex-gpt-5-6-sol` stays admitted as the last fallback. Architecture review stays `claude-fable`, then `pi-glm-5-3-flash-openrouter`.
381
+ `codex-gpt-6-1-sol` is the `reviewer-cli` primary at effort `low`. `pi-qwen3-8-flash-openrouter` is the other-vendor fallback. `codex-gpt-5-6-sol` stays the backup. `codex-gpt-6-1-terra` is the third fallback: harness `codex`, model `gpt-6.1-terra`, vendor `openai`, selector `openai/gpt-6.1-terra`, read-only, effort `low`. OpenRouter listed that id on 2026-11-04 at $1 input and $5 output per 1M tokens through 272,000 prompt tokens ($0.05 cached input, $1.25 cache write), then $2 and $7.50. It is the cost-sensitive Tier 1/2 workhorse in the [workflow catalog](workflow-catalog.md). It does not replace the primary or the backup. Architecture review stays `claude-fable`, then `pi-glm-5-3-flash-openrouter`.
382
382
 
383
383
  Pro is not a separate model. It is `gpt-6.1-sol` with reasoning mode `pro`. OpenRouter lists that mode as `openai/gpt-6.1-sol-pro` at the same per-token price. A route id is `<harness>-<model-slug>`, role `effort` has no `pro` value, and the codex harness ceiling is `reviewer-cli` only. There is no pro architecture-review fallback.
384
384
 
package/docs/reference/workflow-catalog.md CHANGED
@@ -22,7 +22,8 @@ This catalog names common multi-agent workflows by area and lists, for each role
22
22
 
23
23
  The catalog is a four-level taxonomy: **Area → Workflow → Stage → Role**. Each area groups related workflows, each workflow lists its stages in order, and each role belongs to one stage, with a short domain description and up to five ranked candidate models.
24
24
 
25
- - **Candidates are dated research.** Prices (USD per million input and output tokens) and context windows were re-verified against the OpenRouter models list (`https://openrouter.ai/api/v1/models`) on 2026-10-29. Capability notes stay research claims, not guarantees. Verify them again before you dispatch work.
25
+ - **Candidates are dated research.** Prices (USD per million input and output tokens) and context windows were re-verified against the OpenRouter models list (`https://openrouter.ai/api/v1/models`) on 2026-11-04. Capability notes stay research claims, not guarantees. Verify them again before you dispatch work.
26
+ - **2026-11-04 reprice.** `deepseek/deepseek-v4-pro-0813` is $0.115 input and $4.60 output per 1M tokens. That output rate is no longer a Tier 1 workhorse, so those role picks move to `qwen/qwen3.7-plus` ($0.32 / $1.28) or, when Plus is already listed, `google/gemini-3.8-flash` ($0.825 / $4.125). The adversarial security auditor names `moonshotai/kimi-k3.1-pro` instead, because that slot is long security analysis rather than a cheap workhorse. The undated `deepseek/deepseek-v4-pro` id is unchanged at $1.60 / $3.20 and stays a Tier 2 candidate. `deepseek/deepseek-v4-flash-0731` is $0.059 / $0.30. `z-ai/glm-5.3` is $0.08 / $3.60. `z-ai/glm-5.3-flash` is $0.0269 / $0.0996, and that list price is volatile. `moonshotai/kimi-k2.7-code` is $0.60 / $3.50. `mistralai/mistral-large-2512` is still $0.50 / $1.50; the base `mistralai/mistral-large` id ($2.00 / $6.00, 128k context) is not a candidate here. `poolside/laguna-s-2.1` and the xs/free ids are gone from OpenRouter; the TUI slot names `qwen/qwen3.8-flash`. `openai/gpt-6.1-terra` ($1.00 / $5.00 through 272,000 prompt tokens, then $2.00 / $7.50) is the Tier 1/2 cost-sensitive workhorse and the third CLI-critic fallback. `moonshotai/kimi-k3.1-pro` ($2.00 / $10.00) replaces `moonshotai/kimi-k3` on long agentic work, including security and incident follow-through; it is slow and not an interactive model. Live `kimi-k3` is $0.192 / $16.00. `openai/gpt-6-luna` ($0.10 / $0.50) replaces GPT-5.6 Luna in Tier 0. `anthropic/claude-haiku-5.5` ($0.10 / $0.50, 1M context; $0.50 / $2.50 from 100,000 prompt tokens) replaces Haiku 4.5 as the Tier 0 Anthropic backup. `qwen/qwen3-coder-plus` stays absent. The writer roster, the planner (Fable, then Sonnet 5.5), and the CLI critic primary and backup (GPT-6.1 Sol, then GPT-5.6 Sol) are unchanged.
26
27
  - **Writer #2 is Qwen3.7 Plus.** `qwen/qwen3-coder-plus` left the OpenRouter catalog and is not a roster route. Candidate lines that named it name `qwen/qwen3.7-plus`. Re-verified 2026-10-29: $0.32 input and $1.28 output per 1M tokens, 1M context. This repository's writer #2 is the route `pi-qwen3-7-plus-openrouter` (model `openrouter/qwen/qwen3.7-plus`). The writer roster, in order, is `grok-grok-4-7`, `pi-qwen3-7-plus-openrouter`, `agy-gemini-4-argon-high`, `agy-gemini-4-argon-medium`, `agy-gemini-3-8-flash-medium`, `agy-gemini-3-8-flash-high`. [Harness routing](harness-routing.md) is the admission record.
27
28
  - **Removed OpenRouter ids were replaced on 2026-10-29.** `openai/o3-mini-high` is gone. The semantic equivalence verifier now ranks `openai/gpt-6.1-sol` first, the CLI critic in [harness routing](harness-routing.md), and keeps `openai/gpt-5.6-sol` next. `kwaipilot/kat-coder-pro-v2.5` is gone; those slots name `qwen/qwen3-coder`. Catalog roles that named Grok 4.6 now name `x-ai/grok-4.7`. The writer roster is unchanged.
28
29
  - **Candidates are not admission.** A model ID here is a research identifier, not a route selector. Your project admits routes in `.kxm/routes.yaml`, and `kxm harness list` shows which harnesses are installed and signed in. [Harness routing](harness-routing.md) explains when to use a native harness and when an aggregator.
@@ -35,9 +36,9 @@ An interactive `kxm init` offers to install the Software Engineering workflows w
35
36
 
36
37
  | Tier | Price band | Context window | Example models | Typical work |
37
38
  |---|---|---|---|---|
38
- | **Tier 0: Filter and ingestion** | $0.03 - $0.20 / M | 1.0M - 1.05M | DeepSeek-V4-Flash, Qwen-3.7-Flash, GLM-5.3-Flash, GPT-5.6-Luna | High-throughput telemetry, raw log ingestion, triage, video pre-filtering, and initial inbox classification. |
39
- | **Tier 1: Workhorse** | $0.30 - $1.00 / M | 262k - 1.05M | Gemini-3.8-Flash, Qwen-3.7-Plus, DeepSeek-V4-Pro, Devstral-2512 | Routine code generation, multimodal visual QA, Text-to-SQL, AST diff patching, and fast tool dispatching. |
40
- | **Tier 2: Precision and critic** | $1.25 - $4.00 / M | 200k - 1.05M | GPT-5.6-Sol, Grok-4.7, Claude Sonnet 5, OpenAI o3, DeepSeek-R1, Codex 5.3 | Primary code writing, formal contract generation, concurrency race diagnosis, and causal inference. |
39
+ | **Tier 0: Filter and ingestion** | $0.02 - $0.20 / M | 1.0M - 1.05M | DeepSeek-V4-Flash, Qwen-3.7-Flash, GLM-5.3-Flash, GPT-6-Luna, Claude Haiku 5.5 | High-throughput telemetry, raw log ingestion, triage, video pre-filtering, and initial inbox classification. |
40
+ | **Tier 1: Workhorse** | $0.30 - $1.00 / M | 262k - 1.05M | Gemini-3.8-Flash, Qwen-3.7-Plus, GPT-6.1-Terra, Devstral-2512 | Routine code generation, multimodal visual QA, Text-to-SQL, AST diff patching, and fast tool dispatching. |
41
+ | **Tier 2: Precision and critic** | $1.25 - $5.00 / M | 200k - 1.05M | GPT-6.1-Terra, GPT-5.6-Sol, Grok-4.7, Claude Sonnet 5, DeepSeek-V4-Pro, OpenAI o3 | Primary code writing, formal contract generation, concurrency race diagnosis, and causal inference. Terra is the cost-sensitive pick. Sol stays the CLI critic. |
41
42
  | **Tier 3: Architecture** | $5.00 - $15.00 / M | 1.0M - 1.05M | Claude Fable 5.1, Claude Opus 5, GPT-6-Astra-Pro (escalation only) | Architecture review, executive briefs, high-stakes legal redlining, and RFC review. |
42
43
 
43
44
  ---
@@ -132,7 +133,7 @@ This cross-reference points each software and security workflow at the KXM pages
132
133
  *Domain:* Decomposes requirements into an executable, dependency-ordered DAG of modules, tasks, and verification gates.
133
134
 
134
135
  1. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Systems planning; strict boundary isolation and non-overlapping DAG synthesis.
135
- 2. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Exceptional price-to-reasoning ratio; detects race conditions and circular dependencies.
136
+ 2. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Tier 1 price-to-reasoning pick; detects race conditions and circular dependencies.
136
137
  3. **`openai/gpt-6-astra-pro`** (1.05M ctx | $10.00 / $50.00) — Enterprise task decomposition and critical-path scheduling across massive PRDs.
137
138
  4. **`x-ai/grok-4.7`** (500k ctx | $2.00 / $6.00) — Pragmatic engineering focus; rapid modular task decomposition.
138
139
  5. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Broad framework knowledge; maps migration orders and build prerequisites.
@@ -143,11 +144,10 @@ This cross-reference points each software and security workflow at the KXM pages
143
144
 
144
145
  *Domain:* Authors zero-ambiguity API contracts, type definitions, and schema validation rules before implementation begins.
145
146
 
146
- 1. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Premier schema contract generator; 100% compliant OpenAPI 3.1, Protobuf v3, and JSON schemas.
147
- 2. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — High-throughput Zod/Pydantic validation pipeline synthesis.
148
- 3. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Fast translation of data dictionaries into TypeScript interfaces.
149
- 4. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — 1M-context contract verification across monorepo boundaries.
150
- 5. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Reserved for mission-critical authentication and security contract definitions.
147
+ 1. **`openai/gpt-6.1-terra`** (1.05M ctx | $1.00 / $5.00) — Cost-sensitive Tier 1/2 schema workhorse; OpenAPI 3.1, Protobuf v3, and JSON schemas. Above 272,000 prompt tokens the list price is $2.00 / $7.50.
148
+ 2. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Fast translation of data dictionaries into TypeScript interfaces.
149
+ 3. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — 1M-context contract verification across monorepo boundaries.
150
+ 4. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Reserved for mission-critical authentication and security contract definitions.
151
151
 
152
152
  #### Role: Scaffold Build Specialist
153
153
 
@@ -169,7 +169,7 @@ This cross-reference points each software and security workflow at the KXM pages
169
169
 
170
170
  1. **`x-ai/grok-4.7`** (500k ctx | $2.00 / $6.00) — Strong first-pass implementation that passes deterministic verification gates with minimal rework.
171
171
  2. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — 1M-context open-weight anchor; generates performant database layers and CRUD services.
172
- 3. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Exceptional database logic, indexing, and edge-case domain implementation.
172
+ 3. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Database logic, indexing, and edge-case domain implementation.
173
173
  4. **`openai/gpt-5.3-codex`** (400k ctx | $1.75 / $14.00) — Algorithmic precision in Rust, Go, Python, and TypeScript concurrent systems.
174
174
  5. **`qwen/qwen3-coder`** (262k ctx | $0.30 / $1.00) — Multi-file backend logic implementation adhering to project coding patterns.
175
175
 
@@ -191,11 +191,10 @@ This cross-reference points each software and security workflow at the KXM pages
191
191
 
192
192
  *Domain:* Wires external APIs, authentication flows (OAuth2/OIDC), webhook validation, cloud SDKs, and async worker queues.
193
193
 
194
- 1. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Flawless SDK synthesis, webhook signature validation, and API adapter error handling.
194
+ 1. **`openai/gpt-6.1-terra`** (1.05M ctx | $1.00 / $5.00) — Cost-sensitive SDK synthesis, webhook signature validation, and API adapter error handling.
195
195
  2. **`x-ai/grok-4.7`** (500k ctx | $2.00 / $6.00) — Robust glue code, rate-limiting backoff algorithms, and token refresh loops.
196
196
  3. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Ingests full third-party API documentation to emit typed wrapper clients.
197
- 4. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Low-cost mapping of disparate vendor payloads into internal domain models.
198
- 5. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — High-speed webhook listener and protocol transformation adapter.
197
+ 4. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — High-speed webhook listener and protocol transformation adapter.
199
198
 
200
199
  ---
201
200
 
@@ -212,10 +211,10 @@ This cross-reference points each software and security workflow at the KXM pages
212
211
  *Domain:* Analyzes code coupling, designs clean modular boundaries, and authors step-by-step non-breaking migration plans.
213
212
 
214
213
  1. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Strict boundary isolation and dependency inversion planning across large codebases.
215
- 2. **`z-ai/glm-5.3`** (1.05M ctx | $0.0699 / $1.40) — 1.05M context full-repo symbol dependency analysis.
214
+ 2. **`z-ai/glm-5.3`** (1.05M ctx | $0.08 / $3.60) — 1.05M context full-repo symbol dependency analysis.
216
215
  3. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Deep architectural judgment; anticipates runtime blast radius during refactors.
217
- 4. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Cost-efficient full-codebase impact analysis.
218
- 5. **`moonshotai/kimi-k3`** (1.05M ctx | $0.239 / $10.80) — Multi-package refactoring plan generation.
216
+ 4. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Cost-efficient full-codebase impact analysis.
217
+ 5. **`moonshotai/kimi-k3.1-pro`** (1.05M ctx | $2.00 / $10.00) — Multi-package refactoring plan generation. Slow; not an interactive model.
219
218
 
220
219
  #### Role: Code Transform Implementer
221
220
 
@@ -237,7 +236,7 @@ This cross-reference points each software and security workflow at the KXM pages
237
236
 
238
237
  1. **`openai/gpt-6.1-sol`** (1.05M ctx | $2.00 / $10.00) — Deep symbolic verification that AST refactors maintain behavioral equivalence. This is the CLI critic tier in harness routing.
239
238
  2. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Reviews compiler and type-checker cascades against API contracts.
240
- 3. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Type-cascade and semantic regression review across package boundaries.
239
+ 3. **`openai/gpt-6.1-terra`** (1.05M ctx | $1.00 / $5.00) — Third CLI-critic fallback. Cost-sensitive type-cascade and semantic regression review. Primary and backup stay GPT-6.1 Sol and GPT-5.6 Sol.
241
240
  4. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Verifies backwards compatibility and deprecation notices across public APIs.
242
241
  5. **`mistralai/devstral-2512`** (262k ctx | $0.40 / $2.00) — Fast lint rule and syntax compliance review.
243
242
 
@@ -279,7 +278,7 @@ This cross-reference points each software and security workflow at the KXM pages
279
278
  2. **`openai/o3`** (200k ctx | $2.00 / $8.00) — Formal proof of consensus safety, liveness, and split-brain recovery logic.
280
279
  3. **`openai/gpt-6-astra-pro`** (1.05M ctx | $10.00 / $50.00) — Multi-tier cloud topology and global multi-region architecture design.
281
280
  4. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Publication-grade RFC authoring detailing operational trade-offs and team boundaries.
282
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Cost-effective evaluation of microservice communication topologies.
281
+ 5. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Cost-effective evaluation of microservice communication topologies.
283
282
 
284
283
  #### Role: Scalability Database Topologist
285
284
 
@@ -287,7 +286,7 @@ This cross-reference points each software and security workflow at the KXM pages
287
286
 
288
287
  *Domain:* Designs relational/NoSQL schemas, horizontal sharding keys, caching topologies (Redis/Valkey), and zero-downtime migration plans.
289
288
 
290
- 1. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Primary workhorse for high-scale sharding and replication modeling.
289
+ 1. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Primary Tier 1 workhorse for high-scale sharding and replication modeling.
291
290
  2. **`openai/gpt-6-astra-pro`** (1.05M ctx | $10.00 / $50.00) — Distributed multi-master database topology and write-conflict resolution.
292
291
  3. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Strict consistency vs eventual consistency boundary auditing.
293
292
  4. **`qwen/qwen3.8-2.4t-a95b`** (1.05M ctx | $2.00 / $6.00) — Partitioning, indexing, and high-throughput query topology.
@@ -303,7 +302,7 @@ This cross-reference points each software and security workflow at the KXM pages
303
302
  2. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Auth token and protocol threat analysis.
304
303
  3. **`openai/gpt-6-astra-pro`** (1.05M ctx | $10.00 / $50.00) — Enterprise IAM and cloud privilege escalation modeling.
305
304
  4. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — STRIDE threat matrix synthesis and compliance verification.
306
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Full-infrastructure vulnerability surface analysis.
305
+ 5. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Full-infrastructure vulnerability surface analysis.
307
306
 
308
307
  #### Role: IaC Cloud Topology Engineer
309
308
 
@@ -325,9 +324,9 @@ This cross-reference points each software and security workflow at the KXM pages
325
324
 
326
325
  1. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Principal disaster recovery and cascading failure reviewer.
327
326
  2. **`openai/o3`** (200k ctx | $2.00 / $8.00) — Probabilistic modeling of MTTR, MTBF, and network partition recovery.
328
- 3. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Full-topology chaos injection scenario planning.
327
+ 3. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Full-topology chaos injection scenario planning.
329
328
  4. **`openai/gpt-6-astra-pro`** (1.05M ctx | $10.00 / $50.00) — Multi-region disaster recovery runbook formulation.
330
- 5. **`moonshotai/kimi-k3`** (1.05M ctx | $0.239 / $10.80) — Cross-datacenter failover verification.
329
+ 5. **`moonshotai/kimi-k3.1-pro`** (1.05M ctx | $2.00 / $10.00) — Cross-datacenter failover verification. Slow; not an interactive model.
331
330
 
332
331
  #### Role: Architecture Critic Protocol Auditor
333
332
 
@@ -337,7 +336,7 @@ This cross-reference points each software and security workflow at the KXM pages
337
336
 
338
337
  1. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Architecture and permissions criticism.
339
338
  2. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — CLI, protocol and contract criticism.
340
- 3. **`z-ai/glm-5.3`** (1.05M ctx | $0.0699 / $1.40) — 1.05M context full-stack protocol audit.
339
+ 3. **`z-ai/glm-5.3`** (1.05M ctx | $0.08 / $3.60) — 1.05M context full-stack protocol audit.
341
340
  4. **`openai/gpt-6-astra-pro`** (1.05M ctx | $10.00 / $50.00) — Enterprise RFC compliance verification.
342
341
  5. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Independent open-weights architectural validation.
343
342
 
@@ -356,10 +355,10 @@ This cross-reference points each software and security workflow at the KXM pages
356
355
  *Domain:* Read-only audit of developer and operator documentation against CLI `--help` output, source environment variables, tool schemas, and repository configuration before writing begins.
357
356
 
358
357
  1. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Audits authority, permissions, and lifecycle models against actual behavior.
359
- 2. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — High-throughput cross-referencing between source code symbols and markdown documentation.
358
+ 2. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — High-throughput cross-referencing between source code symbols and markdown documentation.
360
359
  3. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Precision auditing of CLI commands, flags, schema definitions, and environment variables.
361
360
  4. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Broad codebase auditing against guides, tutorials, and operational manuals.
362
- 5. **`z-ai/glm-5.3`** (1.05M ctx | $0.0699 / $1.40) — Long-context consistency verification across multi-document doc trees.
361
+ 5. **`z-ai/glm-5.3`** (1.05M ctx | $0.08 / $3.60) — Long-context consistency verification across multi-document doc trees.
363
362
 
364
363
  #### Role: Documentation Writer
365
364
 
@@ -371,7 +370,7 @@ This cross-reference points each software and security workflow at the KXM pages
371
370
  2. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — High-context technical writer; coordinates cross-file references across large documentation sets.
372
371
  3. **`anthropic/claude-sonnet-5`** (1M ctx | $2.00 / $10.00) — Fluid technical documentation authoring with clear instructional hierarchy.
373
372
  4. **`openai/gpt-5.3-codex`** (400k ctx | $1.75 / $14.00) — Accurate technical guides, CLI flag references, and runnable examples.
374
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Cost-effective multi-page documentation generation and reference expansion.
373
+ 5. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Cost-effective multi-page documentation generation and reference expansion.
375
374
 
376
375
  #### Role: Documentation Reviewer
377
376
 
@@ -382,7 +381,7 @@ This cross-reference points each software and security workflow at the KXM pages
382
381
  1. **`anthropic/claude-fable-5.1`** (1M ctx | $10.00 / $50.00) — Verifies security, governance, and authority wording.
383
382
  2. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Verifies CLI command accuracy, flag syntax, and documentation formatting.
384
383
  3. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Rigorous editorial review for clarity, technical completeness, and tone consistency.
385
- 4. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Systematic verification of cross-document links and anchor consistency.
384
+ 4. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Systematic verification of cross-document links and anchor consistency.
386
385
  5. **`mistralai/devstral-2512`** (262k ctx | $0.40 / $2.00) — Rapid lint rule compliance and syntax verification.
387
386
 
388
387
  ---
@@ -447,7 +446,7 @@ This cross-reference points each software and security workflow at the KXM pages
447
446
  1. **`openai/gpt-5.3-codex`** (400k ctx | $1.75 / $14.00) — Native mobile animation drivers, spring physics, and gesture recognizers.
448
447
  2. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Fluid micro-interaction design and gesture state architecture.
449
448
  3. **`x-ai/grok-4.7`** (500k ctx | $2.00 / $6.00) — Fast mobile component implementation and state binding.
450
- 4. **`moonshotai/kimi-k2.7-code`** (262k ctx | $0.56 / $3.10) — Resilient mobile code generation.
449
+ 4. **`moonshotai/kimi-k2.7-code`** (262k ctx | $0.60 / $3.50) — Resilient mobile code generation.
451
450
  5. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Mobile screen layout translation and asset mapping.
452
451
 
453
452
  #### Role: Native Haptics Specialist
@@ -460,7 +459,7 @@ This cross-reference points each software and security workflow at the KXM pages
460
459
  2. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — High-throughput cross-platform haptic wrapper synthesizer.
461
460
  3. **`anthropic/claude-sonnet-5`** (1M ctx | $2.00 / $10.00) — Fine-grained haptic feedback design and event mapping.
462
461
  4. **`openai/gpt-5.4`** (1.05M ctx | $2.50 / $15.00) — Hardware sensor integration and battery lifecycle optimization.
463
- 5. **`openai/gpt-5.6-luna`** (1.05M ctx | $0.20 / $1.20) — Fast AHAP waveform JSON serialization.
462
+ 5. **`openai/gpt-6-luna`** (1.05M ctx | $0.10 / $0.50) — Fast AHAP waveform JSON serialization. Replaces GPT-5.6 Luna in Tier 0.
464
463
 
465
464
  ---
466
465
 
@@ -478,7 +477,7 @@ This cross-reference points each software and security workflow at the KXM pages
478
477
 
479
478
  1. **`x-ai/grok-4.7`** (500k ctx | $2.00 / $6.00) — Pragmatic terminal systems code writer; adheres to strict TUI layout constraints.
480
479
  2. **`openai/gpt-5.3-codex`** (400k ctx | $1.75 / $14.00) — Complex ANSI rendering engines and terminal buffer diffing.
481
- 3. **`poolside/laguna-s-2.1`** (1.05M ctx | $0.09 / $0.18) — Low-cost candidate; 1M context at $0.09/M.
480
+ 3. **`qwen/qwen3.8-flash`** (1M ctx | $0.15 / $0.47) — Low-cost candidate. Laguna S 2.1 left OpenRouter; this 1M-context model is $0.15/M.
482
481
  4. **`mistralai/devstral-2512`** (262k ctx | $0.40 / $2.00) — Clean Go (Bubbletea) and Rust (Ratatui) view/update implementations.
483
482
  5. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — High-ergonomic terminal UX and keyboard navigation design.
484
483
 
@@ -579,7 +578,7 @@ This cross-reference points each software and security workflow at the KXM pages
579
578
  1. **`google/gemini-3.1-pro-preview`** (1.05M ctx | $2.00 / $12.00) — Native multi-image/video reasoning; cross-references character turnarounds across a 1M token context.
580
579
  2. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Real-time frame-by-frame similarity scoring and spatial discrepancy detection.
581
580
  3. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Fine-grained texture, logo, and wardrobe consistency tracking.
582
- 4. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.15 / $0.50) — Ultra-low-cost screening gate across high-volume rendering batches.
581
+ 4. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.0269 / $0.0996) — Ultra-low-cost screening gate across high-volume rendering batches.
583
582
  5. **`openai/gpt-5.4-image-2`** (272k ctx | $8.00 / $15.00) — *Image Output Engine*; generates in-painted corrective reference keyframes.
584
583
 
585
584
  #### Role: Video QA Defect Inspector
@@ -588,7 +587,7 @@ This cross-reference points each software and security workflow at the KXM pages
588
587
 
589
588
  *Domain:* Scans raw video renders for frame flicker, non-physical deformations, temporal morphing, and anatomical glitches.
590
589
 
591
- 1. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.15 / $0.50) — **Tier-0 Primary Ingestion Gate**; 1.05M native video context at $0.15/M prompt. The OpenRouter list price for this id is volatile.
590
+ 1. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.0269 / $0.0996) — **Tier-0 Primary Ingestion Gate**; 1.05M native video context at $0.0269/M prompt. The OpenRouter list price for this id is volatile.
592
591
  2. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Fast, reliable defect classification and timestamp annotation.
593
592
  3. **`google/gemini-3.1-pro-preview`** (1.05M ctx | $2.00 / $12.00) — Escalation arbiter for subtle temporal and physical artifact validation.
594
593
  4. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Evaluates physics continuity (fluid dynamics, shadows, reflections).
@@ -621,10 +620,10 @@ This cross-reference points each software and security workflow at the KXM pages
621
620
  *Domain:* Ingests 1–4 hour recordings, identifying viral hooks, high-retention segments, and standalone topics.
622
621
 
623
622
  1. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Scans multi-hour video/audio streams; extracts timestamped clips at sub-$0.50 cost per hour.
624
- 2. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.15 / $0.50) — 1.05M context for bulk background indexing of livestreams.
623
+ 2. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.0269 / $0.0996) — 1.05M context for bulk background indexing of livestreams.
625
624
  3. **`google/gemini-3.1-pro-preview`** (1.05M ctx | $2.00 / $12.00) — Deep semantic extraction for complex lectures and technical panels.
626
625
  4. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Correlates audience physical cues (laughter, applause) with audio punchlines.
627
- 5. **`moonshotai/kimi-k3`** (1.05M ctx | $0.239 / $10.80) — Long-horizon needle-in-a-haystack topic retrieval across long discussions.
626
+ 5. **`moonshotai/kimi-k3.1-pro`** (1.05M ctx | $2.00 / $10.00) — Long-horizon needle-in-a-haystack topic retrieval across long discussions. Slow; not an interactive model.
628
627
 
629
628
  #### Role: Reframing B-Roll Matcher
630
629
 
@@ -680,7 +679,7 @@ This cross-reference points each software and security workflow at the KXM pages
680
679
  2. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Fast automated gatekeeper in rendering CI/CD; catches black frames and desync.
681
680
  3. **`meta/muse-spark-1.3`** (1.05M ctx | $1.25 / $4.25) — Native audio/video inspection of loudness dynamics and visual pacing.
682
681
  4. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — High spatial resolution checks for color clipping and compression banding.
683
- 5. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.15 / $0.50) — High-throughput preliminary broadcast screening.
682
+ 5. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.0269 / $0.0996) — High-throughput preliminary broadcast screening.
684
683
 
685
684
  ---
686
685
 
@@ -734,7 +733,7 @@ This cross-reference points each software and security workflow at the KXM pages
734
733
 
735
734
  1. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Visual chart layout and structured Vega-Lite/ECharts JSON.
736
735
  2. **`qwen/qwen3.8-flash`** (1M ctx | $0.15 / $0.47) — Ultra-fast, low-cost dashboard widget JSON generation.
737
- 3. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.15 / $0.50) — High-throughput dashboard data series formatting.
736
+ 3. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.0269 / $0.0996) — High-throughput dashboard data series formatting.
738
737
  4. **`anthropic/claude-sonnet-5`** (1M ctx | $2.00 / $10.00) — Executive visual narrative and metric layout design.
739
738
  5. **`openai/gpt-5.6-sol-pro`** (1.05M ctx | $4.00 / $20.00) — Strict schema compliance for BI embedded frameworks.
740
739
 
@@ -756,7 +755,7 @@ This cross-reference points each software and security workflow at the KXM pages
756
755
  2. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — High-capacity Python data science code generator with 1M context.
757
756
  3. **`mistralai/codestral-2508`** (256k ctx | $0.30 / $0.90) — Extremely low-cost Polars/Pandas data pipeline implementation.
758
757
  4. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Complex econometric and Bayesian model formulation.
759
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Large-scale data ingestion and transformation scripts.
758
+ 5. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Large-scale data ingestion and transformation scripts.
760
759
 
761
760
  #### Role: Statistical Causal Reasoner
762
761
 
@@ -788,7 +787,7 @@ This cross-reference points each software and security workflow at the KXM pages
788
787
  2. **`google/gemini-3.1-pro-preview`** (1.05M ctx | $2.00 / $12.00) — Footnote reconciliation and complex accounting table extraction.
789
788
  3. **`qwen/qwen3.8-max-0902`** (1M ctx | $2.00 / $6.00) — Robust bilingual financial reporting and balance sheet analysis.
790
789
  4. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Qualitative MD&A disclosure risk synthesis.
791
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Low-cost long-form structured financial data normalization.
790
+ 5. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Low-cost long-form structured financial data normalization.
792
791
 
793
792
  #### Role: Financial Variance Auditor
794
793
 
@@ -874,7 +873,7 @@ This cross-reference points each software and security workflow at the KXM pages
874
873
  1. **`openai/o3`** (200k ctx | $2.00 / $8.00) — Game-theoretic concession modeling and BATNA optimization.
875
874
  2. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Psychological leverage, interpersonal dynamics, and counter-framing.
876
875
  3. **`deepseek/deepseek-r1-0528`** (164k ctx | $0.50 / $2.15) — Mathematical deal structure and incentive alignment analysis.
877
- 4. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Term sheet covenant structuring.
876
+ 4. **`openai/gpt-6.1-terra`** (1.05M ctx | $1.00 / $5.00) — Cost-sensitive term sheet covenant structuring.
878
877
  5. **`openai/gpt-6-astra`** (1.05M ctx | $10.00 / $50.00) — Complex multi-party merger and joint venture dynamics.
879
878
 
880
879
  ---
@@ -904,7 +903,7 @@ This cross-reference points each software and security workflow at the KXM pages
904
903
  *Domain:* Invokes external APIs (calendar, email, CRM), coordinates webhooks, and manages background task pipelines.
905
904
 
906
905
  1. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Sub-second deterministic tool dispatch with zero argument hallucination.
907
- 2. **`openai/gpt-5.6-sol`** (1.05M ctx | $4.00 / $20.00) — Complex multi-step tool dependency chaining and error recovery.
906
+ 2. **`openai/gpt-6.1-terra`** (1.05M ctx | $1.00 / $5.00) — Cost-sensitive multi-step tool dependency chaining and error recovery.
908
907
  3. **`qwen/qwen3.6-flash`** (1M ctx | $0.1875 / $1.125) — High-throughput sub-second personal action router.
909
908
  4. **`anthropic/claude-sonnet-5`** (1M ctx | $2.00 / $10.00) — Ambiguous user intent disambiguation.
910
909
  5. **`google/gemini-3.1-pro-preview-customtools`** (1.05M ctx | $2.00 / $12.00) — Multi-agent tool orchestration.
@@ -915,11 +914,11 @@ This cross-reference points each software and security workflow at the KXM pages
915
914
 
916
915
  *Domain:* Classifies incoming emails, extracts action items, prioritizes blockers, and drafts context-aware replies.
917
916
 
918
- 1. **`deepseek/deepseek-v4-flash-0731`** (1.05M ctx | $0.119 / $0.238) — High-volume Tier-0 spam/priority email triage at $0.119/M.
919
- 2. **`openai/gpt-5.6-luna`** (1.05M ctx | $0.20 / $1.20) — Fast structured JSON action item extractor from emails.
917
+ 1. **`deepseek/deepseek-v4-flash-0731`** (1.05M ctx | $0.059 / $0.30) — High-volume Tier-0 spam/priority email triage at $0.059/M.
918
+ 2. **`openai/gpt-6-luna`** (1.05M ctx | $0.10 / $0.50) — Fast structured JSON action item extractor from emails.
920
919
  3. **`google/gemini-3.5-flash-lite`** (1.05M ctx | $0.30 / $2.50) — Multimodal attachment and PDF receipt processing.
921
920
  4. **`mistralai/mistral-small-3.2-24b-instruct`** (256k ctx | $0.075 / $0.20) — Private, fast thread summarization.
922
- 5. **`anthropic/claude-haiku-4.5`** (200k ctx | $1.00 / $5.00) — Escalation tier for sensitive VIP relationship messages.
921
+ 5. **`anthropic/claude-haiku-5.5`** (1M ctx | $0.10 / $0.50) — Tier 0 Anthropic backup for sensitive VIP relationship messages. Base rate; prompts at or above 100,000 tokens list at $0.50 / $2.50.
923
922
 
924
923
  #### Role: Executive Briefing Concierge
925
924
 
@@ -999,8 +998,8 @@ This cross-reference points each software and security workflow at the KXM pages
999
998
  *Domain:* Ingests high-velocity JSON logs, OpenTelemetry traces, and time-series metrics at rock-bottom token cost.
1000
999
 
1001
1000
  1. **`qwen/qwen3.7-flash`** (1M ctx | $0.03 / $0.13) — **Tier-0 Primary Streaming Gate**; 1M context at $0.03/M prompt.
1002
- 2. **`deepseek/deepseek-v4-flash-0731`** (1.05M ctx | $0.119 / $0.238) — High-throughput anomaly classification in 1.05M context.
1003
- 3. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.15 / $0.50) — High-speed log aggregation and pattern recognition.
1001
+ 2. **`deepseek/deepseek-v4-flash-0731`** (1.05M ctx | $0.059 / $0.30) — High-throughput anomaly classification in 1.05M context.
1002
+ 3. **`z-ai/glm-5.3-flash`** (1.05M ctx | $0.0269 / $0.0996) — High-speed log aggregation and pattern recognition.
1004
1003
  4. **`google/gemini-3.1-flash-lite`** (1.05M ctx | $0.25 / $1.50) — Multimodal metric graph and trace ingestion.
1005
1004
  5. **`amazon/nova-2-lite-v1`** (1M ctx | $0.30 / $2.50) — Resilient secondary cloud telemetry classifier.
1006
1005
 
@@ -1010,9 +1009,9 @@ This cross-reference points each software and security workflow at the KXM pages
1010
1009
 
1011
1010
  *Domain:* Consumes raw multi-gigabyte log dumps, APM spans, and core dumps; filters noise and isolates fault boundaries.
1012
1011
 
1013
- 1. **`deepseek/deepseek-v4-flash-0731`** (1.05M ctx | $0.119 / $0.238) — **Tier-0 Primary Ingestion Gate**; 1.05M context at $0.119/M prompt.
1012
+ 1. **`deepseek/deepseek-v4-flash-0731`** (1.05M ctx | $0.059 / $0.30) — **Tier-0 Primary Ingestion Gate**; 1.05M context at $0.059/M prompt.
1014
1013
  2. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Multimodal crash analysis (logs + APM graphs + error screenshots).
1015
- 3. **`openai/gpt-5.6-luna`** (1.05M ctx | $0.20 / $1.20) — Fast structured JSON classification of stack traces across 1M tokens.
1014
+ 3. **`openai/gpt-6-luna`** (1.05M ctx | $0.10 / $0.50) — Fast structured JSON classification of stack traces across 1M tokens.
1016
1015
  4. **`x-ai/grok-4.20`** (2M ctx | $1.25 / $2.50) — 2M context window for cross-service distributed tracing scans.
1017
1016
  5. **`qwen/qwen3.8-flash`** (1M ctx | $0.15 / $0.47) — High-throughput multilingual log scanner for multi-cluster events.
1018
1017
 
@@ -1050,7 +1049,7 @@ This cross-reference points each software and security workflow at the KXM pages
1050
1049
  2. **`anthropic/claude-sonnet-5`** (1M ctx | $2.00 / $10.00) — High-velocity technical post-mortem report writer.
1051
1050
  3. **`x-ai/grok-4.7`** (500k ctx | $2.00 / $6.00) — Root cause extraction without corporate sugarcoating.
1052
1051
  4. **`google/gemini-3.8-flash`** (1.05M ctx | $0.825 / $4.125) — Visual timeline generation combining metric spikes and log traces.
1053
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Infrastructure remediations and preventive patch specification.
1052
+ 5. **`qwen/qwen3.7-plus`** (1M ctx | $0.32 / $1.28) — Infrastructure remediations and preventive patch specification.
1054
1053
 
1055
1054
  ---
1056
1055
 
@@ -1082,7 +1081,7 @@ This cross-reference points each software and security workflow at the KXM pages
1082
1081
  2. **`openai/o3`** (200k ctx | $2.00 / $8.00) — Fuzzing parameter generation and mathematical boundary mutation testing.
1083
1082
  3. **`anthropic/claude-opus-5`** (1M ctx | $5.00 / $25.00) — Unforgiving security critic; catches incomplete sanitization and secondary injections.
1084
1083
  4. **`nvidia/nemotron-3-ultra-550b-a55b`** (262k ctx | $0.60 / $2.40) — 550B MoE frontier open reasoning model for vulnerability search.
1085
- 5. **`deepseek/deepseek-v4-pro-0813`** (1.05M ctx | $0.66 / $1.98) — Full-codebase taint tracking and dataflow analysis.
1084
+ 5. **`moonshotai/kimi-k3.1-pro`** (1.05M ctx | $2.00 / $10.00) — Full-codebase taint tracking and dataflow analysis. Slow; not an interactive model.
1086
1085
 
1087
1086
  ---
1088
1087
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@kontextmind/kxm",
3
- "version": "0.7.209",
3
+ "version": "0.7.211",
4
4
  "description": "KXM local-first multi-agent orchestration and operator dashboard",
5
5
  "type": "module",
6
6
  "author": "KontextMind",
package/plugins/kxm/.claude-plugin/plugin.json CHANGED
@@ -2,7 +2,7 @@
2
2
  "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
3
3
  "name": "kxm",
4
4
  "displayName": "KXM",
5
- "version": "0.7.209",
5
+ "version": "0.7.211",
6
6
  "description": "Headless multi-agent orchestration, durable workflows, and a live operator dashboard for Pi and Claude Code",
7
7
  "author": {
8
8
  "name": "KontextMind",
package/plugins/kxm/dist/cli.js CHANGED
@@ -57070,6 +57070,7 @@ var ADMITTED_GUIDE_BINDINGS = Object.freeze({
57070
57070
  "anthropic/fable": { harness: "claude", provider: "anthropic", model: "fable" },
57071
57071
  "openai/gpt-6.1-sol": { harness: "codex", provider: "openai", model: "gpt-6.1-sol" },
57072
57072
  "openai/gpt-5.6-sol": { harness: "codex", provider: "openai", model: "gpt-5.6-sol" },
57073
+ "openai/gpt-6.1-terra": { harness: "codex", provider: "openai", model: "gpt-6.1-terra" },
57073
57074
  "x-ai/grok-4.7": { harness: "grok", provider: "xai", model: "grok-4.7" },
57074
57075
  "xai/grok-4.7": { harness: "grok", provider: "xai", model: "grok-4.7" },
57075
57076
  "qwen/qwen3.7-plus": { harness: "pi", provider: "openrouter", model: "qwen/qwen3.7-plus" }
@@ -57088,7 +57089,7 @@ var GUIDE_WORKFLOWS = [
57088
57089
  domain: "Decomposes requirements into an executable, dependency-ordered DAG of modules, tasks, and verification gates.",
57089
57090
  candidates: [
57090
57091
  { vendor: "anthropic", model: "claude-fable-5.1" },
57091
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57092
+ { vendor: "qwen", model: "qwen3.7-plus" },
57092
57093
  { vendor: "openai", model: "gpt-6-astra-pro" },
57093
57094
  { vendor: "x-ai", model: "grok-4.7" },
57094
57095
  { vendor: "qwen", model: "qwen3.8-max-0902" }
@@ -57100,8 +57101,7 @@ var GUIDE_WORKFLOWS = [
57100
57101
  role: "spec-contract-generator",
57101
57102
  domain: "Authors zero-ambiguity API contracts, type definitions, and schema validation rules before implementation begins.",
57102
57103
  candidates: [
57103
- { vendor: "openai", model: "gpt-5.6-sol" },
57104
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57104
+ { vendor: "openai", model: "gpt-6.1-terra" },
57105
57105
  { vendor: "google", model: "gemini-3.8-flash" },
57106
57106
  { vendor: "qwen", model: "qwen3.7-plus" },
57107
57107
  { vendor: "anthropic", model: "claude-fable-5.1" }
@@ -57128,7 +57128,7 @@ var GUIDE_WORKFLOWS = [
57128
57128
  candidates: [
57129
57129
  { vendor: "x-ai", model: "grok-4.7" },
57130
57130
  { vendor: "qwen", model: "qwen3.7-plus" },
57131
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57131
+ { vendor: "google", model: "gemini-3.8-flash" },
57132
57132
  { vendor: "openai", model: "gpt-5.3-codex" },
57133
57133
  { vendor: "qwen", model: "qwen3-coder" }
57134
57134
  ]
@@ -57152,10 +57152,9 @@ var GUIDE_WORKFLOWS = [
57152
57152
  role: "sdk-integration-specialist",
57153
57153
  domain: "Wires external APIs, authentication flows (OAuth2/OIDC), webhook validation, cloud SDKs, and async worker queues.",
57154
57154
  candidates: [
57155
- { vendor: "openai", model: "gpt-5.6-sol" },
57155
+ { vendor: "openai", model: "gpt-6.1-terra" },
57156
57156
  { vendor: "x-ai", model: "grok-4.7" },
57157
57157
  { vendor: "qwen", model: "qwen3.7-plus" },
57158
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57159
57158
  { vendor: "google", model: "gemini-3.8-flash" }
57160
57159
  ]
57161
57160
  }
@@ -57176,8 +57175,8 @@ var GUIDE_WORKFLOWS = [
57176
57175
  { vendor: "anthropic", model: "claude-fable-5.1" },
57177
57176
  { vendor: "z-ai", model: "glm-5.3" },
57178
57177
  { vendor: "anthropic", model: "claude-opus-5" },
57179
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57180
- { vendor: "moonshotai", model: "kimi-k3" }
57178
+ { vendor: "qwen", model: "qwen3.7-plus" },
57179
+ { vendor: "moonshotai", model: "kimi-k3.1-pro" }
57181
57180
  ]
57182
57181
  },
57183
57182
  {
@@ -57201,7 +57200,7 @@ var GUIDE_WORKFLOWS = [
57201
57200
  candidates: [
57202
57201
  { vendor: "openai", model: "gpt-6.1-sol" },
57203
57202
  { vendor: "openai", model: "gpt-5.6-sol" },
57204
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57203
+ { vendor: "openai", model: "gpt-6.1-terra" },
57205
57204
  { vendor: "anthropic", model: "claude-opus-5" },
57206
57205
  { vendor: "mistralai", model: "devstral-2512" }
57207
57206
  ]
@@ -57245,7 +57244,7 @@ var GUIDE_WORKFLOWS = [
57245
57244
  { vendor: "openai", model: "o3" },
57246
57245
  { vendor: "openai", model: "gpt-6-astra-pro" },
57247
57246
  { vendor: "anthropic", model: "claude-opus-5" },
57248
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" }
57247
+ { vendor: "qwen", model: "qwen3.7-plus" }
57249
57248
  ]
57250
57249
  },
57251
57250
  {
@@ -57254,7 +57253,7 @@ var GUIDE_WORKFLOWS = [
57254
57253
  role: "scalability-database-topologist",
57255
57254
  domain: "Designs relational/NoSQL schemas, horizontal sharding keys, caching topologies, and zero-downtime migration plans.",
57256
57255
  candidates: [
57257
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57256
+ { vendor: "qwen", model: "qwen3.7-plus" },
57258
57257
  { vendor: "openai", model: "gpt-6-astra-pro" },
57259
57258
  { vendor: "anthropic", model: "claude-fable-5.1" },
57260
57259
  { vendor: "qwen", model: "qwen3.8-2.4t-a95b" },
@@ -57271,7 +57270,7 @@ var GUIDE_WORKFLOWS = [
57271
57270
  { vendor: "openai", model: "gpt-5.6-sol" },
57272
57271
  { vendor: "openai", model: "gpt-6-astra-pro" },
57273
57272
  { vendor: "anthropic", model: "claude-opus-5" },
57274
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" }
57273
+ { vendor: "qwen", model: "qwen3.7-plus" }
57275
57274
  ]
57276
57275
  },
57277
57276
  {
@@ -57295,9 +57294,9 @@ var GUIDE_WORKFLOWS = [
57295
57294
  candidates: [
57296
57295
  { vendor: "anthropic", model: "claude-fable-5.1" },
57297
57296
  { vendor: "openai", model: "o3" },
57298
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57297
+ { vendor: "qwen", model: "qwen3.7-plus" },
57299
57298
  { vendor: "openai", model: "gpt-6-astra-pro" },
57300
- { vendor: "moonshotai", model: "kimi-k3" }
57299
+ { vendor: "moonshotai", model: "kimi-k3.1-pro" }
57301
57300
  ]
57302
57301
  },
57303
57302
  {
@@ -57328,7 +57327,7 @@ var GUIDE_WORKFLOWS = [
57328
57327
  domain: "Read-only audit of developer and operator documentation against CLI --help outputs, source environment variables, tool schemas, and repository configuration.",
57329
57328
  candidates: [
57330
57329
  { vendor: "anthropic", model: "claude-fable-5.1" },
57331
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57330
+ { vendor: "qwen", model: "qwen3.7-plus" },
57332
57331
  { vendor: "openai", model: "gpt-5.6-sol" },
57333
57332
  { vendor: "qwen", model: "qwen3.8-max-0902" },
57334
57333
  { vendor: "z-ai", model: "glm-5.3" }
@@ -57344,7 +57343,7 @@ var GUIDE_WORKFLOWS = [
57344
57343
  { vendor: "qwen", model: "qwen3.7-plus" },
57345
57344
  { vendor: "anthropic", model: "claude-sonnet-5" },
57346
57345
  { vendor: "openai", model: "gpt-5.3-codex" },
57347
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" }
57346
+ { vendor: "google", model: "gemini-3.8-flash" }
57348
57347
  ]
57349
57348
  },
57350
57349
  {
@@ -57356,7 +57355,7 @@ var GUIDE_WORKFLOWS = [
57356
57355
  { vendor: "anthropic", model: "claude-fable-5.1" },
57357
57356
  { vendor: "openai", model: "gpt-5.6-sol" },
57358
57357
  { vendor: "anthropic", model: "claude-opus-5" },
57359
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
57358
+ { vendor: "qwen", model: "qwen3.7-plus" },
57360
57359
  { vendor: "mistralai", model: "devstral-2512" }
57361
57360
  ]
57362
57361
  }
package/plugins/kxm/dist/mcp-server.js CHANGED
@@ -25462,7 +25462,7 @@ function sessionTokenFixHint(policy) {
25462
25462
  }
25463
25463
 
25464
25464
  // plugins/kxm/src/mcp-server.ts
25465
- var VERSION = "0.7.209";
25465
+ var VERSION = "0.7.211";
25466
25466
  var CONFIGURE_PLUGIN = "/plugin configure kxm@kxm";
25467
25467
  var inbox = /* @__PURE__ */ new Map();
25468
25468
  var notifiedInbox = /* @__PURE__ */ new Set();
package/plugins/kxm/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "claude-plugin",
3
- "version": "0.7.209",
3
+ "version": "0.7.211",
4
4
  "private": true,
5
5
  "type": "module",
6
6
  "engines": {
package/plugins/kxm/src/init-guide-setup.ts CHANGED
@@ -63,6 +63,7 @@ const ADMITTED_GUIDE_BINDINGS: Readonly<Record<string, AgentBinding>> = Object.f
63
63
  "anthropic/fable": { harness: "claude", provider: "anthropic", model: "fable" },
64
64
  "openai/gpt-6.1-sol": { harness: "codex", provider: "openai", model: "gpt-6.1-sol" },
65
65
  "openai/gpt-5.6-sol": { harness: "codex", provider: "openai", model: "gpt-5.6-sol" },
66
+ "openai/gpt-6.1-terra": { harness: "codex", provider: "openai", model: "gpt-6.1-terra" },
66
67
  "x-ai/grok-4.7": { harness: "grok", provider: "xai", model: "grok-4.7" },
67
68
  "xai/grok-4.7": { harness: "grok", provider: "xai", model: "grok-4.7" },
68
69
  "qwen/qwen3.7-plus": { harness: "pi", provider: "openrouter", model: "qwen/qwen3.7-plus" },
@@ -83,7 +84,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
83
84
  domain: "Decomposes requirements into an executable, dependency-ordered DAG of modules, tasks, and verification gates.",
84
85
  candidates: [
85
86
  { vendor: "anthropic", model: "claude-fable-5.1" },
86
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
87
+ { vendor: "qwen", model: "qwen3.7-plus" },
87
88
  { vendor: "openai", model: "gpt-6-astra-pro" },
88
89
  { vendor: "x-ai", model: "grok-4.7" },
89
90
  { vendor: "qwen", model: "qwen3.8-max-0902" },
@@ -95,8 +96,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
95
96
  role: "spec-contract-generator",
96
97
  domain: "Authors zero-ambiguity API contracts, type definitions, and schema validation rules before implementation begins.",
97
98
  candidates: [
98
- { vendor: "openai", model: "gpt-5.6-sol" },
99
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
99
+ { vendor: "openai", model: "gpt-6.1-terra" },
100
100
  { vendor: "google", model: "gemini-3.8-flash" },
101
101
  { vendor: "qwen", model: "qwen3.7-plus" },
102
102
  { vendor: "anthropic", model: "claude-fable-5.1" },
@@ -123,7 +123,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
123
123
  candidates: [
124
124
  { vendor: "x-ai", model: "grok-4.7" },
125
125
  { vendor: "qwen", model: "qwen3.7-plus" },
126
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
126
+ { vendor: "google", model: "gemini-3.8-flash" },
127
127
  { vendor: "openai", model: "gpt-5.3-codex" },
128
128
  { vendor: "qwen", model: "qwen3-coder" },
129
129
  ],
@@ -147,10 +147,9 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
147
147
  role: "sdk-integration-specialist",
148
148
  domain: "Wires external APIs, authentication flows (OAuth2/OIDC), webhook validation, cloud SDKs, and async worker queues.",
149
149
  candidates: [
150
- { vendor: "openai", model: "gpt-5.6-sol" },
150
+ { vendor: "openai", model: "gpt-6.1-terra" },
151
151
  { vendor: "x-ai", model: "grok-4.7" },
152
152
  { vendor: "qwen", model: "qwen3.7-plus" },
153
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
154
153
  { vendor: "google", model: "gemini-3.8-flash" },
155
154
  ],
156
155
  },
@@ -171,8 +170,8 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
171
170
  { vendor: "anthropic", model: "claude-fable-5.1" },
172
171
  { vendor: "z-ai", model: "glm-5.3" },
173
172
  { vendor: "anthropic", model: "claude-opus-5" },
174
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
175
- { vendor: "moonshotai", model: "kimi-k3" },
173
+ { vendor: "qwen", model: "qwen3.7-plus" },
174
+ { vendor: "moonshotai", model: "kimi-k3.1-pro" },
176
175
  ],
177
176
  },
178
177
  {
@@ -196,7 +195,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
196
195
  candidates: [
197
196
  { vendor: "openai", model: "gpt-6.1-sol" },
198
197
  { vendor: "openai", model: "gpt-5.6-sol" },
199
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
198
+ { vendor: "openai", model: "gpt-6.1-terra" },
200
199
  { vendor: "anthropic", model: "claude-opus-5" },
201
200
  { vendor: "mistralai", model: "devstral-2512" },
202
201
  ],
@@ -240,7 +239,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
240
239
  { vendor: "openai", model: "o3" },
241
240
  { vendor: "openai", model: "gpt-6-astra-pro" },
242
241
  { vendor: "anthropic", model: "claude-opus-5" },
243
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
242
+ { vendor: "qwen", model: "qwen3.7-plus" },
244
243
  ],
245
244
  },
246
245
  {
@@ -249,7 +248,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
249
248
  role: "scalability-database-topologist",
250
249
  domain: "Designs relational/NoSQL schemas, horizontal sharding keys, caching topologies, and zero-downtime migration plans.",
251
250
  candidates: [
252
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
251
+ { vendor: "qwen", model: "qwen3.7-plus" },
253
252
  { vendor: "openai", model: "gpt-6-astra-pro" },
254
253
  { vendor: "anthropic", model: "claude-fable-5.1" },
255
254
  { vendor: "qwen", model: "qwen3.8-2.4t-a95b" },
@@ -266,7 +265,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
266
265
  { vendor: "openai", model: "gpt-5.6-sol" },
267
266
  { vendor: "openai", model: "gpt-6-astra-pro" },
268
267
  { vendor: "anthropic", model: "claude-opus-5" },
269
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
268
+ { vendor: "qwen", model: "qwen3.7-plus" },
270
269
  ],
271
270
  },
272
271
  {
@@ -290,9 +289,9 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
290
289
  candidates: [
291
290
  { vendor: "anthropic", model: "claude-fable-5.1" },
292
291
  { vendor: "openai", model: "o3" },
293
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
292
+ { vendor: "qwen", model: "qwen3.7-plus" },
294
293
  { vendor: "openai", model: "gpt-6-astra-pro" },
295
- { vendor: "moonshotai", model: "kimi-k3" },
294
+ { vendor: "moonshotai", model: "kimi-k3.1-pro" },
296
295
  ],
297
296
  },
298
297
  {
@@ -323,7 +322,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
323
322
  domain: "Read-only audit of developer and operator documentation against CLI --help outputs, source environment variables, tool schemas, and repository configuration.",
324
323
  candidates: [
325
324
  { vendor: "anthropic", model: "claude-fable-5.1" },
326
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
325
+ { vendor: "qwen", model: "qwen3.7-plus" },
327
326
  { vendor: "openai", model: "gpt-5.6-sol" },
328
327
  { vendor: "qwen", model: "qwen3.8-max-0902" },
329
328
  { vendor: "z-ai", model: "glm-5.3" },
@@ -339,7 +338,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
339
338
  { vendor: "qwen", model: "qwen3.7-plus" },
340
339
  { vendor: "anthropic", model: "claude-sonnet-5" },
341
340
  { vendor: "openai", model: "gpt-5.3-codex" },
342
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
341
+ { vendor: "google", model: "gemini-3.8-flash" },
343
342
  ],
344
343
  },
345
344
  {
@@ -351,7 +350,7 @@ export const GUIDE_WORKFLOWS: readonly GuideWorkflow[] = [
351
350
  { vendor: "anthropic", model: "claude-fable-5.1" },
352
351
  { vendor: "openai", model: "gpt-5.6-sol" },
353
352
  { vendor: "anthropic", model: "claude-opus-5" },
354
- { vendor: "deepseek", model: "deepseek-v4-pro-0813" },
353
+ { vendor: "qwen", model: "qwen3.7-plus" },
355
354
  { vendor: "mistralai", model: "devstral-2512" },
356
355
  ],
357
356
  },
package/plugins/kxm/src/mcp-server.ts CHANGED
@@ -11,7 +11,7 @@ import { deliverInboxNotification } from "./inbox.ts";
11
11
  import type { HubEvent, MessageRecord } from "./protocol.ts";
12
12
  import { sessionTokenFixHint } from "./session-token-hint.ts";
13
13
 
14
- const VERSION = "0.7.209";
14
+ const VERSION = "0.7.211";
15
15
  const CONFIGURE_PLUGIN = "/plugin configure kxm@kxm";
16
16
  const inbox = new Map<string, MessageRecord>();
17
17
  const notifiedInbox = new Set<string>();
package/scripts/harness-run.mjs CHANGED
@@ -164,7 +164,7 @@ export const ROUTES = Object.freeze({
164
164
  provider: "openai",
165
165
  roles: Object.freeze(["reviewer-cli"]),
166
166
  permissions: Object.freeze(["read-only"]),
167
- models: Object.freeze(["gpt-6.1-sol", "gpt-5.6-sol"]),
167
+ models: Object.freeze(["gpt-6.1-sol", "gpt-5.6-sol", "gpt-6.1-terra"]),
168
168
  efforts: Object.freeze(["low", "medium", "high"]),
169
169
  auth: Object.freeze({ args: ["login", "status"], loginHint: "codex login" }),
170
170
  },