job-forge 2.14.11 → 2.14.13

package/iso/agents/general-free.md

@@ -1,24 +1,16 @@
 ---
-description: Procedural worker on free-tier model. Use for form filling via Geometra, tracker updates, TSV merges, scan dedup, OTP retrieval, and other mechanical/scripted tasks where quality-sensitive text generation is NOT required.
+description: Procedural worker on the low-cost DeepSeek V4 Flash OpenCode route. Use for form filling via Geometra, tracker updates, TSV merges, scan dedup, OTP retrieval, and other mechanical/scripted tasks where quality-sensitive text generation is NOT required.
 role: fast
 targets:
-  # No inline model: iso-route's "standard" preset maps role "fast" to
-  # each harness's cheapest reliable model. Claude Code reads
+  # No inline model: JobForge's models.yaml maps role "fast" to each
+  # harness's cheapest reliable model. On OpenCode that is pinned to
+  # opencode-go/deepseek-v4-flash. Claude Code reads
   # .claude/iso-route.resolved.json; OpenCode reads opencode.json's
   # agent.fast.model (iso-harness 0.6.0+).
   opencode:
     mode: subagent
     temperature: 0.1
     reasoningEffort: minimal
-    # Primary comes from models.yaml: opencode/big-pickle on OpenCode.
-    # Fallback chain stays free-only and intentionally excludes
-    # openrouter/minimax/minimax-m2.5:free because recent traces showed
-    # repeated read({ path|file_path }) schema drift on that route.
-    fallback_models:
-      - openrouter/z-ai/glm-4.5-air:free
-      - openrouter/openai/gpt-oss-20b:free
-      - openrouter/nvidia/nemotron-3-nano-30b-a3b:free
-      - openrouter/qwen/qwen3-coder:free
     tools:
       geometra_connect: true
       geometra_page_model: true
@@ -34,7 +26,7 @@ targets:
       task: false
 ---
 
-You are the @general-free subagent. You run on a free-tier model, which means the orchestrator has delegated this task to you **specifically because the work is procedural**: deterministic steps, scripted outputs, no nuanced writing required.
+You are the @general-free subagent. You run on JobForge's low-cost procedural model, which means the orchestrator has delegated this task to you **specifically because the work is procedural**: deterministic steps, scripted outputs, no nuanced writing required.
 
 ## Run This Pre-Flight First Every Time
 

package/iso/agents/general-paid.md

@@ -1,28 +1,16 @@
 ---
-description: Quality-sensitive worker on the strongest free-tier OpenCode model by default. Use for offer evaluation narratives (Blocks A-F), cover letter generation, "Why X?" form answers, interview STAR stories, and other tasks where writing quality and judgment matter.
+description: Quality-sensitive worker on the low-cost DeepSeek V4 Flash OpenCode route by default. Use for offer evaluation narratives (Blocks A-F), cover letter generation, "Why X?" form answers, interview STAR stories, and other tasks where writing quality and judgment matter.
 role: quality
 targets:
-  # No inline model: JobForge's models.yaml maps role "quality" to a
-  # free OpenRouter model on OpenCode, while Claude/Codex keep their
-  # quality-tier defaults from the standard preset. Claude Code reads
+  # No inline model: JobForge's models.yaml maps role "quality" to
+  # opencode-go/deepseek-v4-flash on OpenCode, while Claude/Codex keep
+  # their quality-tier defaults from the standard preset. Claude Code reads
   # .claude/iso-route.resolved.json; OpenCode reads opencode.json's
   # agent.quality.model (iso-harness 0.6.0+).
   opencode:
     mode: subagent
     temperature: 0.3
     reasoningEffort: medium
-    # Primary (qwen/qwen3-next-80b-a3b-instruct:free) resolves from the
-    # openrouter-free preset. First fallbacks intentionally avoid another
-    # immediate hop through the same Venice/Qwen pool when OpenRouter
-    # returns "[Venice] insufficient …" — gpt-oss-120b + nemotron are
-    # usually different backends. Remaining picks stay free-only.
-    fallback_models:
-      - openrouter/openai/gpt-oss-120b:free
-      - openrouter/nvidia/nemotron-3-super-120b-a12b:free
-      - openrouter/z-ai/glm-4.5-air:free
-      - openrouter/qwen/qwen3-coder:free
-      - openrouter/google/gemma-4-31b-it:free
-      - openrouter/meta-llama/llama-3.3-70b-instruct:free
     tools:
       geometra_connect: true
       geometra_page_model: true
@@ -40,10 +28,10 @@ targets:
 
 You are the @general-paid subagent. The orchestrator delegated this task to you because it requires quality writing or judgment — the kind of work `@general-free` isn't well-suited for.
 
-On OpenCode, this agent now defaults to a free OpenRouter model. On other
-harnesses, the same role may still resolve to a premium model. Your job is
-still the same: produce the best final writing you can from the context you
-were given.
+On OpenCode, this agent defaults to DeepSeek V4 Flash so application work
+does not fall back into overloaded free OpenRouter pools. On other harnesses,
+the same role may still resolve to a premium model. Your job is still the
+same: produce the best final writing you can from the context you were given.
 
 ## Do These Tasks
 

package/iso/agents/glm-minimal.md

@@ -1,23 +1,16 @@
 ---
-description: Narrow-scope extractor on free-tier model. Use for single-purpose tasks where the orchestrator passes the exact input and expects a small, structured output — e.g., "extract these 8 fields from this JD text" or "parse this form schema into a label→type map". NOT for multi-step workflows.
+description: Narrow-scope extractor on the low-cost DeepSeek V4 Flash OpenCode route. Use for single-purpose tasks where the orchestrator passes the exact input and expects a small, structured output — e.g., "extract these 8 fields from this JD text" or "parse this form schema into a label→type map". NOT for multi-step workflows.
 role: minimal
 targets:
-  # No inline model: iso-route's "standard" preset maps role "minimal" to
-  # each harness's smallest credible model. Claude Code reads
+  # No inline model: JobForge's models.yaml maps role "minimal" to each
+  # harness's smallest credible model. On OpenCode that is pinned to
+  # opencode-go/deepseek-v4-flash. Claude Code reads
   # .claude/iso-route.resolved.json; OpenCode reads opencode.json's
   # agent.minimal.model (iso-harness 0.6.0+).
   opencode:
     mode: subagent
     temperature: 0
     reasoningEffort: none
-    # Primary (openai/gpt-oss-20b:free) resolves from openrouter-free
-    # preset. Fallback chain sticks to small dense models with reliable
-    # structured-output behavior — no creative generation upstream.
-    fallback_models:
-      - openrouter/google/gemma-4-26b-a4b-it:free
-      - openrouter/nvidia/nemotron-nano-9b-v2:free
-      - openrouter/google/gemma-4-31b-it:free
-      - openrouter/z-ai/glm-4.5-air:free
     tools:
       geometra_*: false
       gmail_*: false

package/iso/commands/job-forge.md

@@ -140,13 +140,18 @@ When the user says "apply to N jobs", "process the pipeline", or similar, execut
 
 ```
 Step 1  — Enumerate candidates
-  - Grep data/applications/$(date +%Y-%m-%d).md and the last 3 day files for status "Evaluated"
+  - Grep data/applications/*.md for status "Evaluated" without loading every file into context
   - Also read data/pipeline.md for unprocessed URLs
   - Build ordered list: candidates = [job_1, job_2, ..., job_N]
 
 Step 2  — Dedup against already-applied
-  - For each candidate, Grep data/pipeline.md + today's day file for "APPLIED" + company+role
-  - Drop any match. Never re-apply.
+  - For each candidate, grep all four sources for URL and company+role:
+    data/pipeline.md, data/applications/*.md, batch/tracker-additions/*.tsv,
+    batch/tracker-additions/merged/*.tsv
+  - Drop any APPLIED / Applied match before counting toward N. Never re-apply.
+  - If a subagent later returns SKIP because it found a duplicate, treat that as
+    a missed preflight check; finish the current round, re-run dedupe, then pick
+    a replacement from the remaining candidates.
 
 Step 3  — Pre-flight cleanup (once, before the loop)
   - geometra_list_sessions()

package/iso/config.json

@@ -1,49 +1,8 @@
 {
-  "opencodeModelFallback": {
-    "cooldown_seconds": 60,
-    "timeout_seconds": 30,
-    "notify_on_fallback": true,
-    "fallback_models": [
-      "openrouter/openai/gpt-oss-120b:free",
-      "openrouter/z-ai/glm-4.5-air:free",
-      "openrouter/nvidia/nemotron-3-super-120b-a12b:free",
-      "openrouter/qwen/qwen3-next-80b-a3b-instruct:free"
-    ],
-    "retryable_error_patterns": [
-      "\\bvenice\\b",
-      "insufficient\\s+usd",
-      "insufficient\\s+.*\\s+diem",
-      "diem\\s+balance",
-      "add\\s+credits",
-      "chutes",
-      "insufficient\\s+(?:credits?|funds?|balance)",
-      "credit.*balance.*too.*low",
-      "(?:temporarily\\s+)?unavailable|overloaded|try\\s+again"
-    ]
-  },
   "targets": {
     "opencode": {
-      "plugin": ["@razroo/opencode-model-fallback"],
       "instructions": ["templates/states.yml"],
-      "small_model": "openrouter/google/gemma-4-26b-a4b-it:free",
-      "provider": {
-        "openrouter": {
-          "models": {
-            "qwen/qwen3-coder:free": {},
-            "z-ai/glm-4.5-air:free": {},
-            "qwen/qwen3-next-80b-a3b-instruct:free": {},
-            "openai/gpt-oss-20b:free": {},
-            "openai/gpt-oss-120b:free": {},
-            "minimax/minimax-m2.5:free": {},
-            "nvidia/nemotron-3-super-120b-a12b:free": {},
-            "nvidia/nemotron-3-nano-30b-a3b:free": {},
-            "nvidia/nemotron-nano-9b-v2:free": {},
-            "google/gemma-4-26b-a4b-it:free": {},
-            "google/gemma-4-31b-it:free": {},
-            "meta-llama/llama-3.3-70b-instruct:free": {}
-          }
-        }
-      }
+      "small_model": "opencode-go/deepseek-v4-flash"
     }
   }
 }

package/iso/instructions.md

@@ -5,9 +5,9 @@ AI-powered job search pipeline: scans portals, evaluates offers, generates CVs v
 ## Hard limits
 
 - [H1] Max 2 parallel `task` dispatches per message. For N jobs, run `ceil(N/2)` sequential rounds of 2. A round is not complete until both subagents return a final outcome (`APPLIED`, `APPLY FAILED`, `SKIP`, `Discarded`, or a written TSV path). A `task` tool result that only gives a session id / title is a launch acknowledgement, not completion. Applies in all modes, for all user phrasings ("urgent", "apply to 10 jobs now").
-  why: higher parallelism blows through free-tier rate limits; each subagent requires post-cleanup and racing more than 2 reliably loses at least one result. On 2026-04-25 the orchestrator launched round 2 while round 1 had only returned task ids, leaving four application subagents in flight and losing two provider-fallback recoveries
+  why: each subagent requires post-cleanup and racing more than 2 reliably loses at least one result. On 2026-04-25 the orchestrator launched round 2 while round 1 had only returned task ids, leaving four application subagents in flight and losing two provider recoveries
 
-- [H2] Max 1 application per company+role. Before every `apply` dispatch, grep all four sources for the URL and for `company+role`: `data/pipeline.md`, all `data/applications/*.md` day files, `batch/tracker-additions/*.tsv`, `batch/tracker-additions/merged/*.tsv`. If any source shows APPLIED / Applied, skip the dispatch.
+- [H2] Max 1 application per company+role. Before every `apply` dispatch, grep all four sources for the URL and for `company+role`: `data/pipeline.md`, all `data/applications/*.md` day files, `batch/tracker-additions/*.tsv`, `batch/tracker-additions/merged/*.tsv`. If any source shows APPLIED / Applied, skip the dispatch and pick a replacement from the remaining candidate list. Do not count duplicates toward a requested "apply to N jobs" total, and do not delegate obvious duplicates just so a subagent can return SKIP.
   why: 2026-04 same-day batch collision — when two batches target the same role, `npx job-forge merge` updates the existing day-file row rather than appending, so grepping day files alone misses earlier-batch applies; merged/*.tsv is the only place the breadcrumb remains
 
 - [H3] Before every batch of `task` dispatches that will use Geometra, call `geometra_list_sessions` then `geometra_disconnect({closeBrowser: true})`. Every round, no exceptions. Name this cleanup as an explicit "step 0" in your first-response plan for any multi-apply request — it is the most frequently skipped guardrail in practice, and skipping it produces cascade "Not connected" failures on the next dispatch.
@@ -37,7 +37,7 @@ AI-powered job search pipeline: scans portals, evaluates offers, generates CVs v
   why: iso-trace showed 0.25% Agent calls across 5174 turns under a prior over-broad "delegate before 2nd tool call" rule — the rule was ignored in practice; narrowing matches the original cache-bust incident
 
 - [D2] Route subagent work by cost tier. `@general-free`: procedural — form-fill, TSV merge, verify, OTP retrieval, portal scan metadata extraction, one-shot structured-field transforms. `@general-paid`: quality-sensitive — offer evaluation narrative Blocks A-F, cover letters, "Why X?" answers, STAR interview stories, LinkedIn outreach. `@glm-minimal`: narrow ≤5K-input one-shot extract/classify jobs that do not need context.
-  why: GLM 5.1 doesn't discount cache reads so procedural work there costs ~10×; free-tier models handle procedural work fine empirically (`opencode/big-pickle` processed 1000+ messages at $0)
+  why: OpenCode routes all JobForge tiers through DeepSeek V4 Flash by default now; recent traces showed free OpenRouter fallbacks freezing or hitting provider balance errors during applications
 
 - [D3] Read the active mode file before dispatch. Mode files own score gates, provider fallback, portal runbooks, and output shape.
   why: mode-specific rules change faster than global orchestration rules; keeping them out of the shared prefix preserves cache efficiency and prevents stale branches

package/models.yaml

@@ -1,17 +1,15 @@
 # JobForge model policy.
 #
-# Extends @razroo/iso-route's bundled "openrouter-free" preset — a
-# curated zero-cost routing config that keeps Claude Code / Codex on
-# their native providers but routes OpenCode through explicit free
-# OpenRouter model IDs. Preset content lives in
-# node_modules/@razroo/iso-route/presets/openrouter-free.yaml; run
-# `npx iso-route plan models.yaml` to see the resolved policy.
+# Extends @razroo/iso-route's bundled "standard" preset, then pins every
+# OpenCode route to DeepSeek V4 Flash. Recent traces showed free OpenRouter
+# routes freezing or falling through Venice balance errors, so JobForge's
+# OpenCode default is now "best affordable paid" rather than "free".
 #
 # JobForge's subagents bind to preset roles via the `role:` field in
 # iso/agents/<slug>.md:
-#   @general-free  → role: fast     (Haiku / OpenCode big-pickle / gpt-5.4-mini)
-#   @general-paid  → role: quality  (Opus 4.7 / OpenRouter Qwen3 Next 80B free / gpt-5.4)
-#   @glm-minimal   → role: minimal  (Haiku / OpenRouter GPT-OSS-20B free / gpt-5.4-nano)
+#   @general-free  → role: fast     (Haiku / DeepSeek V4 Flash / gpt-5.4-mini)
+#   @general-paid  → role: quality  (Opus 4.7 / DeepSeek V4 Flash / gpt-5.4)
+#   @glm-minimal   → role: minimal  (Haiku / DeepSeek V4 Flash / gpt-5.4-nano)
 #
 # Override anything by adding fields here. For example, to pin Opus on
 # Claude Code for the @general-paid (quality) role:
@@ -29,11 +27,27 @@
 #           provider: openai
 #           model: gpt-5.4
 
-extends: openrouter-free
+extends: standard
+
+default:
+  targets:
+    opencode:
+      provider: opencode
+      model: opencode-go/deepseek-v4-flash
 
 roles:
   fast:
     targets:
       opencode:
         provider: opencode
-        model: opencode/big-pickle
+        model: opencode-go/deepseek-v4-flash
+  quality:
+    targets:
+      opencode:
+        provider: opencode
+        model: opencode-go/deepseek-v4-flash
+  minimal:
+    targets:
+      opencode:
+        provider: opencode
+        model: opencode-go/deepseek-v4-flash

package/modes/apply.md

@@ -17,7 +17,7 @@ Live application assistant. Reads the active application form in Chrome (via Geo
   why: prior aborted subagents leave Chromium sessions stuck in the pool; next `geometra_connect` fails with "Not connected" (see root `[H3]`)
 
 - [H5] Max 2 parallel `task` dispatches per round. For N jobs, run `ceil(N/2)` sequential rounds of 2. Never emit 3+ dispatches in a single message. Do not start the next round until both current-round subagents return final outcomes (`APPLIED`, `APPLY FAILED`, `SKIP`, `Discarded`, or a written TSV path); task/session ids are only launch receipts.
-  why: free-tier rate limits + subagent post-cleanup cost; racing more than 2 reliably loses at least one result (see root `[H1]`). A 2026-04-25 OpenCode trace launched round 2 while round 1 was still running, then lost two fallback recoveries
+  why: subagent post-cleanup cost and portal state make racing more than 2 unreliable (see root `[H1]`). A 2026-04-25 OpenCode trace launched round 2 while round 1 was still running, then lost two provider recoveries
 
 ## Defaults
 
@@ -46,10 +46,10 @@ Live application assistant. Reads the active application form in Chrome (via Geo
   why: class-B Ashby / Cloudflare-fronted portals need a residential outbound IP; the fix is wired in Geometra MCP v1.59.0 but the orchestrator owns the config pipe. See "BYO Residential Proxy" in modes/reference-portals.md.
 
 - [D8] Upgrade application routing to `@general-paid` when the offer score is ≥ 4.0/5, the user flags "top-tier", "dream job", or "high-stakes", or the candidate is late-stage/post-screen.
-  why: form-fill flows are 6+ steps; free-tier sometimes aborts mid-flow on large Greenhouse/Workday schemas; paid tier has more headroom
+  why: high-stakes applications need the quality-sensitive prompt and medium reasoning budget even though OpenCode now routes both application tiers through DeepSeek V4 Flash by default
 
-- [D9] If an upgraded `@general-paid` subagent fails with provider-side errors, re-dispatch the same URL once on `@general-free` before marking FAILED. Provider-side errors include Venice, Diem, Chutes, HTTP 402/429, insufficient credits/funds/balance, overload, and temporarily unavailable.
-  why: OpenCode paid-tier routing can still use free OpenRouter model IDs; backend pool limits are not evidence that a procedural free-tier worker cannot complete the same form after preflight gates pass
+- [D9] If a subagent fails with provider-side errors, do not auto-downgrade or re-dispatch the same URL. Report the provider failure, leave any TSV untouched unless there is a confirmed outcome, and inspect telemetry before retrying.
+  why: OpenCode now pins all JobForge application tiers to DeepSeek V4 Flash; switching `@general-paid` → `@general-free` changes the prompt/tool budget but not the provider route, so automatic duplicate dispatches add risk without fixing provider availability
 
 ## Procedure
 
@@ -65,7 +65,7 @@ Live application assistant. Reads the active application form in Chrome (via Geo
 10. Generate answers from Block B + Block F + Section G + JD.
 11. Submit as ONE `run_actions` call [H1] using labels [D6] with `imeFriendly: true` [D4].
 12. On session error, run the 4-step recovery; only one retry [H2].
-13. On upgraded-provider failure, downgrade once to `@general-free` [D9].
+13. On provider failure, stop and inspect telemetry before any retry [D9].
 14. On OTP prompt, fetch the code from Gmail via `gmail_get_message`.
 15. Submit the OTP with `geometra_fill_otp` and click Submit.
 16. Write outcome as `batch/tracker-additions/*.tsv` [H3].
@@ -98,7 +98,7 @@ Or, on failure:
 APPLY FAILED AFTER RECOVERY: <url>
   Error 1: <first error>
   Error 2: <post-recovery error>
-  Recommend: re-dispatch on @general-paid
+  Recommend: inspect telemetry before retrying this URL
 ```
 
 ---
@@ -176,7 +176,10 @@ When `location_constraints` is absent, use the prose fields:
 
 ```
 Step 1  — Build the job list (N items)
-Step 2  — Dedup: Grep data/pipeline.md + today's day file for each company+role. Drop any already APPLIED.
+Step 2  — Dedup: for each candidate, grep all four sources for the URL and for company+role:
+          data/pipeline.md, all data/applications/*.md day files,
+          batch/tracker-additions/*.tsv, batch/tracker-additions/merged/*.tsv.
+          Drop any already APPLIED before counting toward N; pick replacements from the remaining list.
 Step 3  — geometra_list_sessions() + geometra_disconnect({closeBrowser: true})  [once, before loop]
 Step 4  — For round in ceil(N/2):
             pair = jobs[round*2 : round*2 + 2]
@@ -192,7 +195,7 @@ Step 6  — Reconcile outcomes (Hard Limit #6):
 Step 7  — Summarize outcomes; do NOT auto-retry failures.
 ```
 
-If a subagent fails, report it in the summary and let the user decide whether to retry. Never auto-retry — re-running a submit step risks duplicate applications.
+If a subagent fails, report it in the summary and let the user decide whether to retry. Never auto-retry — re-running a submit step risks duplicate applications. If a subagent returns SKIP because it discovered a duplicate, treat that as a missed preflight check: finish the current round, then choose a replacement candidate only after re-running dedupe against all four sources.
 
 **Outcome routing (Hard Limit #6 in `AGENTS.md`):**
 - Subagents write `batch/tracker-additions/{num}-{slug}.tsv` — one TSV per job.
@@ -359,10 +362,10 @@ Call 4:  geometra_run_actions({
    APPLY FAILED AFTER RECOVERY: <URL>
    Error 1: <first error message>
    Error 2: <error after recovery>
-   Recommend: re-dispatch on @general-paid
+   Recommend: inspect telemetry before retrying this URL
    ```
 
-   Do NOT try a third time. Do NOT try a different approach. The orchestrator will decide whether to re-dispatch on a bigger model.
+   Do NOT try a third time. Do NOT try a different approach. The orchestrator will decide whether to retry after inspecting telemetry.
 
 ### Skip schema re-fetches mid-flow (Rule D)
 

package/opencode.json

@@ -1,33 +1,15 @@
 {
   "$schema": "https://opencode.ai/config.json",
-  "model": "openrouter/qwen/qwen3-coder:free",
+  "model": "opencode-go/deepseek-v4-flash",
   "agent": {
     "fast": {
-      "model": "opencode/big-pickle"
+      "model": "opencode-go/deepseek-v4-flash"
     },
     "quality": {
-      "model": "openrouter/qwen/qwen3-next-80b-a3b-instruct:free"
+      "model": "opencode-go/deepseek-v4-flash"
     },
     "minimal": {
-      "model": "openrouter/openai/gpt-oss-20b:free"
-    }
-  },
-  "provider": {
-    "openrouter": {
-      "models": {
-        "qwen/qwen3-coder:free": {},
-        "z-ai/glm-4.5-air:free": {},
-        "qwen/qwen3-next-80b-a3b-instruct:free": {},
-        "openai/gpt-oss-20b:free": {},
-        "openai/gpt-oss-120b:free": {},
-        "minimax/minimax-m2.5:free": {},
-        "nvidia/nemotron-3-super-120b-a12b:free": {},
-        "nvidia/nemotron-3-nano-30b-a3b:free": {},
-        "nvidia/nemotron-nano-9b-v2:free": {},
-        "google/gemma-4-26b-a4b-it:free": {},
-        "google/gemma-4-31b-it:free": {},
-        "meta-llama/llama-3.3-70b-instruct:free": {}
-      }
+      "model": "opencode-go/deepseek-v4-flash"
     }
   },
   "mcp": {
@@ -66,11 +48,8 @@
       }
     }
   },
-  "plugin": [
-    "@razroo/opencode-model-fallback"
-  ],
   "instructions": [
     "templates/states.yml"
   ],
-  "small_model": "openrouter/google/gemma-4-26b-a4b-it:free"
+  "small_model": "opencode-go/deepseek-v4-flash"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "job-forge",
-  "version": "2.14.11",
+  "version": "2.14.13",
   "description": "AI-powered job search pipeline built on opencode",
   "type": "module",
   "bin": {
@@ -97,7 +97,6 @@
     "@razroo/iso": "^0.2.5",
     "@razroo/iso-eval": "^0.4.0",
     "@razroo/iso-harness": "^0.6.1",
-    "@razroo/iso-route": "^0.5.3",
-    "@razroo/opencode-model-fallback": "^0.3.1"
+    "@razroo/iso-route": "^0.5.3"
   }
 }

package/scripts/check-iso-smoke.mjs

@@ -21,9 +21,9 @@ const checks = [
   ["H7 distrusts subagent prose", () => every(files.instructions, ["must originate from a file", "not from prior subagent prose"])],
   ["shared prompt points to on-demand references", () => every(files.instructions, ["modes/{mode}.md", "modes/reference-setup.md", "modes/reference-portals.md", "modes/reference-geometra.md"])],
   ["apply mode owns high-stakes upgrade", () => every(files.apply, ["[D8]", "@general-paid", "4.0/5", "high-stakes"])],
-  ["apply mode owns provider downgrade", () => every(files.apply, ["[D9]", "@general-free", "HTTP 402/429", "insufficient credits/funds/balance"])],
-  ["models policy extends free OpenRouter preset", () => /extends:\s*openrouter-free/.test(files.models)],
-  ["OpenCode fallback plugin is configured", () => every(files.config, ["opencodeModelFallback", "@razroo/opencode-model-fallback"])],
+  ["apply mode blocks provider auto-downgrade", () => every(files.apply, ["[D9]", "do not auto-downgrade", "inspect telemetry before retrying"])],
+  ["models policy pins OpenCode to DeepSeek V4 Flash", () => /extends:\s*standard/.test(files.models) && count(files.models, "opencode-go/deepseek-v4-flash") >= 4],
+  ["OpenCode fallback plugin is not configured", () => !every(files.config, ["opencodeModelFallback", "@razroo/opencode-model-fallback"])],
 ];
 
 const failures = checks
@@ -41,3 +41,7 @@ console.log(`JobForge iso smoke passed (${checks.length} checks).`);
 function every(source, needles) {
   return needles.every((needle) => source.includes(needle));
 }
+
+function count(source, needle) {
+  return source.split(needle).length - 1;
+}