[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7FD1FCMQSMBQFQ59Y74MMJ1":3,"skill-same:01M7FD1FCMQSMBQFQ59Y74MMJ1":46},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7FD1FCMQSMBQFQ59Y74MMJ1","update-model-versions",null,"Check whether newer versions of the AI models we already use (fal.ai image, video\u002Fmotion, audio; BytePlus Ark ids; OpenRouter text) have shipped, and open a PR bumping any genuine successor. Use when asked to \"check for model updates\", \"are our models current\", \"bump models\", or when run by the daily model-freshness routine. Only bumps EXISTING models to a newer version of the same model — it does not add net-new models. npm dependency bumps (including @tanstack\u002Fai*) are out of scope: Dependabot owns those.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fupdate-model-versions\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: update-model-versions\ndescription: >\n  Check whether newer versions of the AI models we already use (fal.ai image,\n  video\u002Fmotion, audio; BytePlus Ark ids; OpenRouter text) have shipped, and\n  open a PR bumping any genuine successor. Use when asked to \"check for model\n  updates\", \"are our models current\", \"bump models\", or when run by the daily\n  model-freshness routine. Only bumps EXISTING models to a newer version of\n  the same model — it does not add net-new models. npm dependency bumps\n  (including @tanstack\u002Fai*) are out of scope: Dependabot owns those.\n---\n\n# Update model versions\n\nOur model registries are the single source of truth:\n\n| Class                   | File                          | Export                   |\n| ----------------------- | ----------------------------- | ------------------------ |\n| Text (OpenRouter)       | `src\u002Fmodels\u002Fmodels.config.ts` | `SCRIPT_ANALYSIS_MODELS` |\n| Image (fal.ai)          | `src\u002Fmodels\u002Fmodels.ts`        | `IMAGE_MODELS`           |\n| Video \u002F motion (fal.ai) | `src\u002Fmodels\u002Fmodels.ts`        | `IMAGE_TO_VIDEO_MODELS`  |\n| Audio (fal.ai)          | `src\u002Fmodels\u002Fmodels.ts`        | `AUDIO_MODELS`           |\n| BytePlus Ark ids        | `src\u002Fmodels\u002Fmodels.ts`        | `byteplusId` fields      |\n\nA model with a native BytePlus route (#1157) carries **two ids under one\nkey**: the fal endpoint `id` AND a `byteplusId` (Ark model id). A version bump\nmust move BOTH — bumping only the fal id silently routes Ark traffic to the\nolder model (and vice versa), and the pricing alias\n(`applyBytePlusRouteAliases`) would quote the wrong rate for whichever id was\nleft behind.\n\nThe goal each run: detect newer versions → verify each is a real successor →\nopen a focused PR that bumps it → leave everything green.\n\n**npm dependencies are out of scope.** Bumping `@tanstack\u002Fai*` (or any other npm\npackage) is Dependabot's job, not this routine's — the detector no longer checks\nthe npm registry. Don't open model-freshness PRs or issues for package versions.\n\n## 1. Detect candidates\n\n```bash\nbun models:check          # human-readable report\nbun models:check --json   # { ok, errorCount, hasUpdates, models[] }\n```\n\n`scripts\u002Fcheck-model-updates.ts` reads the registries and queries public,\nunauthenticated catalogs (fal.ai `\u002Fapi\u002Fmodels`, OpenRouter `\u002Fapi\u002Fv1\u002Fmodels`).\nBytePlus Ark ids are checked **offline** against the installed\n`@tanstack\u002Fai-byteplus` model catalog — Ark publishes no unauthenticated\ncatalog, so the adapter's lists are the freshness source. Everything else is\nHTTP-only so it runs anywhere — no `FAL_KEY` or MCP needed.\nBehind a proxy it routes through `curl` (Bun's fetch can't traverse a\nTLS-intercepting proxy), so `curl` must be on PATH in that case.\n\nCandidates are **heuristic** (same brand, higher version number, same modality,\nnot already adopted). Treat them as leads to verify, not facts.\n\n**Check `ok` before trusting the result.** `ok: false` (equivalently, a non-zero\nexit code or `errorCount > 0`) means one or more lookups FAILED — the report is\nINCOMPLETE, not \"all current\". Do not treat a failed run as \"nothing to do\":\nfix connectivity and re-run. Only when `ok` is `true` does `hasUpdates: false`\ngenuinely mean every model is current — in that case, stop.\n\n## 2. Verify each candidate is a genuine successor\n\nDo not bump on the heuristic alone. For each flagged candidate decide: _is this\nthe same model, one version newer — or a different product line \u002F tier?_\n\n**fal models — prefer the fal tooling (richest signal):**\n\n- **genmedia CLI** (what the fal community skills use; works headless):\n  ```bash\n  curl https:\u002F\u002Fgenmedia.sh\u002Finstall -fsS | bash   # once, if missing\n  genmedia setup --non-interactive --api-key \"$FAL_KEY\"\n  genmedia models --endpoint_id \u003Ccandidate-id> --json   # confirm it exists\n  genmedia schema  \u003Ccandidate-id> --json                # compare input params\n  genmedia pricing \u003Ccandidate-id> --json                # cost delta\n  ```\n- **fal-ai MCP** if available in this session: `search_models`,\n  `get_model_schema`, `get_pricing` give the same data.\n- **Fallback (zero-auth):** the OpenAPI spec — a 200 means the endpoint is real:\n  `curl -s -o \u002Fdev\u002Fnull -w \"%{http_code}\" \"https:\u002F\u002Ffal.ai\u002Fapi\u002Fopenapi\u002Fqueue\u002Fopenapi.json?endpoint_id=\u003Cid>\"`,\n  and `https:\u002F\u002Ffal.ai\u002Fmodels\u002F\u003Cpath>\u002Fllms.txt` for param specs (see CLAUDE.md).\n\n**text models:** confirm the candidate id resolves on\n`https:\u002F\u002Fopenrouter.ai\u002Fapi\u002Fv1\u002Fmodels` and keeps the same tier (don't turn a\n`-mini` into a non-mini, or a `pro` into `flash`).\n\n**BytePlus Ark ids:** verify against the live ModelArk docs\n(`docs.byteplus.com\u002Fen\u002Fdocs\u002FModelArk\u002F…`) plus the installed adapter's\n`model-meta` (its JSDoc records doc-vs-reality findings, e.g. per-model\nresolution tiers). Ark ids end in a `yymmdd` snapshot date; tier suffixes\n(`-fast`, `-mini`, `-lite`, `-pro`) are different products, not successors.\nTwo Ark-specific gotchas: a new model answers **404 ModelNotOpen until\nactivated in the Ark console** (verify at request time, not startup), and\nresolution tiers are per-model AND per-tier priced — the adapter's model-meta\nand BytePlus's own pricing page have disagreed about which tiers exist (e.g.\nSeedance 2.5's 1080p), so check the official rate table\n(`docs.byteplus.com\u002Fen\u002Fdocs\u002FModelArk\u002F1544106`) for the tier the request\nbuilders ask for, and confirm the rate-card entry matches THAT tier's price.\n\n**Reject a candidate when** it is a different tier\u002Fvariant (fast\u002Flite\u002Fstandard\nvs pro, lora\u002Ftrainer\u002Fedit gear), a different modality, a preview\u002Fexperimental\nbuild replacing a stable one, or its schema\u002Fpricing changed so much it needs\nproduct judgement. When unsure, skip it and note it in the PR body rather than\nguessing.\n\n## 3. Apply the bump (one PR per upgrade)\n\nWork one upgrade at a time so each PR is reviewable and revertible.\n\n**Idempotency — check first:** `gh pr list --state open --search \"in:title model\"`.\nIf an open PR or branch already covers this exact bump, skip it. Branch name:\n`auto\u002Fmodel-update-\u003Cregistry-key>-\u003Cnew-version>` (e.g.\n`auto\u002Fmodel-update-minimax_hailuo_02-2.3`).\n\n**Golden rule: change the `id` (and metadata), never the registry KEY.** Keys\nare persisted in the DB (a team's selected model) and referenced across selectors\nand schemas — renaming one is a breaking migration, out of scope here.\n\nPer class, edit and follow through:\n\n- **Text** (`models.config.ts`): update `id`, `name`, `description`,\n  `contextWindow`. If you bump `DEFAULT_ANALYSIS_MODEL`'s model, the constant\n  references a key, so it's unaffected.\n  **Catalog-lag bridge:** the `@tanstack\u002Fai-openrouter` adapter ships a codegen\n  snapshot of OpenRouter's model list that lags new releases, so a freshly\n  shipped id may not be in its typed union yet. When a text-model bump adopts an\n  id the installed catalog lacks, `bun typecheck` fails at the `createAdapter`\n  call sites — add a `createModel` entry for the id to `CATALOG_LAG_MODELS`\n  (`src\u002Fmodels\u002Fserver\u002Fcreate-adapter.ts`) with the correct `input` modalities, plus\n  `features: ['reasoning', 'structured_outputs']` when the model supports them.\n  (The reverse — pruning a bridged id once the adapter package catches up — is\n  handled by Dependabot's package bump, guided by `catalog-lag.test.ts`, not by\n  this routine.)\n- **Image** (`models.ts` `IMAGE_MODELS`): update `id`, `name`, `description`,\n  `maxPromptLength`. If the model has an `EDIT_ENDPOINTS` entry, update that\n  endpoint id too. Confirm the new endpoint still supports the edit\u002Freference\n  flow if it had one.\n- **Video \u002F motion** (`models.ts` `IMAGE_TO_VIDEO_MODELS`): update `id` + meta,\n  then regenerate the schemas — **`bun motion:codegen`** (writes\n  `src\u002Fmotion\u002Fserver\u002Fgenerated\u002F**` and `src\u002Fmotion\u002Fserver\u002Fendpoint-map.ts`). Never hand-write motion\n  schemas. Re-check `maxPromptLength` against the new schema.\n- **Audio** (`models.ts` `AUDIO_MODELS`): update `id` + `capabilities`\n  (durations, formats) from the new schema.\n- **BytePlus-routed models** (a `byteplusId` on the entry): bump the\n  `byteplusId` in the same PR as the fal id — never one without the other.\n  Then follow through:\n  - `src\u002Fbilling\u002Fbyteplus-pricing.ts` — the rate card is keyed by Ark model id,\n    so the old id's entry must be replaced with the new id at the new model's\n    advertised rate (BytePlus publishes no pricing API; read the pricing page,\n    re-date the header comment, and note the rate is advertised-not-verified).\n    A missed rename means Ark generations bill $0 — exactly the #1069 failure\n    mode the card exists to prevent.\n  - `src\u002Fbilling\u002Ffal-cost.ts` `ENDPOINT_STRATEGY` — if the old fal endpoint ids\n    appear there (token-billed Seedance endpoints do), rename them too.\n  - The request builders (`build-byteplus-video-request.ts` \u002F\n    `build-byteplus-image-request.ts`) read the id from the registry, but\n    re-check their baked-in assumptions (resolution tier, watermark default,\n    frame-vs-reference roles) against the new model's docs.\n  - e2e fixtures live under a fal-endpoint-keyed dir and each fixture's\n    `match.model` is the fal endpoint id, which aimock matches on — migrate\n    them (rename dir + edit `match.model`) rather than re-recording when the\n    prompt is unchanged.\n- **fal pricing:** model ids are pricing keys in `model_pricing` (D1, see `src\u002Fbilling\u002Fserver\u002Ffal-pricing-live.ts`)\n  (auto-generated). After any fal id change run **`bun scripts\u002Fupdate-fal-pricing.ts`**\n  (needs `FAL_KEY`). If it can't run, add the new id's pricing manually via the\n  override path documented in that script and flag it in the PR.\n\n## 4. Quality gates (must pass before opening the PR)\n\n```bash\nbun typecheck\nbun lint\nbun run test src\u002Fmodels src\u002Fmotion   # registry + motion suites\nbun run test src\u002Fbilling             # pricing\u002Fcost if fal pricing changed\n```\n\nFix anything that breaks. If a bump cascades into non-trivial changes (schema\nshape changed, pricing model differs, tests need rework), stop and open an\n**issue** describing it instead of forcing a half-working PR.\n\n## 5. Open the PR\n\n```bash\ngit checkout -b auto\u002Fmodel-update-\u003Ckey>-\u003Cversion>\ngit commit -am \"chore(models): bump \u003Cname> \u003Cold> → \u003Cnew>\"\ngh pr create --title \"chore(models): bump \u003Cname> \u003Cold> → \u003Cnew>\" --body \"\u003Cbody>\"\n```\n\nPR body must include, per change: registry key, old id → new id, why it's a\ngenuine successor, links (fal model page \u002F OpenRouter), pricing delta, and\nwhich generated files \u002F pricing were regenerated. End with: \"🤖 Opened by the\ndaily model-freshness routine (#792). Verify pricing & quality before merge.\"\n\nLeave PRs as **draft** only if a quality gate is amber (e.g. pricing couldn't be\nauto-fetched); otherwise ready-for-review. Do not merge — a human reviews.\n","# Update model versions\n\nOur model registries are the single source of truth:\n\n| Class                   | File                          | Export                   |\n| ----------------------- | ----------------------------- | ------------------------ |\n| Text (OpenRouter)       | `src\u002Fmodels\u002Fmodels.config.ts` | `SCRIPT_ANALYSIS_MODELS` |\n| Image (fal.ai)          | `src\u002Fmodels\u002Fmodels.ts`        | `IMAGE_MODELS`           |\n| Video \u002F motion (fal.ai) | `src\u002Fmodels\u002Fmodels.ts`        | `IMAGE_TO_VIDEO_MODELS`  |\n| Audio (fal.ai)          | `src\u002Fmodels\u002Fmodels.ts`        | `AUDIO_MODELS`           |\n| BytePlus Ark ids        | `src\u002Fmodels\u002Fmodels.ts`        | `byteplusId` fields      |\n\nA model with a native BytePlus route (#1157) carries **two ids under one\nkey**: the fal endpoint `id` AND a `byteplusId` (Ark model id). A version bump\nmust move BOTH — bumping only the fal id silently routes Ark traffic to the\nolder model (and vice versa), and the pricing alias\n(`applyBytePlusRouteAliases`) would quote the wrong rate for whichever id was\nleft behind.\n\nThe goal each run: detect newer versions → verify each is a real successor →\nopen a focused PR that bumps it → leave everything green.\n\n**npm dependencies are out of scope.** Bumping `@tanstack\u002Fai*` (or any other npm\npackage) is Dependabot's job, not this routine's — the detector no longer checks\nthe npm registry. Don't open model-freshness PRs or issues for package versions.\n\n## 1. Detect candidates\n\n```bash\nbun models:check          # human-readable report\nbun models:check --json   # { ok, errorCount, hasUpdates, models[] }\n```\n\n`scripts\u002Fcheck-model-updates.ts` reads the registries and queries public,\nunauthenticated catalogs (fal.ai `\u002Fapi\u002Fmodels`, OpenRouter `\u002Fapi\u002Fv1\u002Fmodels`).\nBytePlus Ark ids are checked **offline** against the installed\n`@tanstack\u002Fai-byteplus` model catalog — Ark publishes no unauthenticated\ncatalog, so the adapter's lists are the freshness source. Everything else is\nHTTP-only so it runs anywhere — no `FAL_KEY` or MCP needed.\nBehind a proxy it routes through `curl` (Bun's fetch can't traverse a\nTLS-intercepting proxy), so `curl` must be on PATH in that case.\n\nCandidates are **heuristic** (same brand, higher version number, same modality,\nnot already adopted). Treat them as leads to verify, not facts.\n\n**Check `ok` before trusting the result.** `ok: false` (equivalently, a non-zero\nexit code or `errorCount > 0`) means one or more lookups FAILED — the report is\nINCOMPLETE, not \"all current\". Do not treat a failed run as \"nothing to do\":\nfix connectivity and re-run. Only when `ok` is `true` does `hasUpdates: false`\ngenuinely mean every model is current — in that case, stop.\n\n## 2. Verify each candidate is a genuine successor\n\nDo not bump on the heuristic alone. For each flagged candidate decide: _is this\nthe same model, one version newer — or a different product line \u002F tier?_\n\n**fal models — prefer the fal tooling (richest signal):**\n\n- **genmedia CLI** (what the fal community skills use; works headless):\n  ```bash\n  curl https:\u002F\u002Fgenmedia.sh\u002Finstall -fsS | bash   # once, if missing\n  genmedia setup --non-interactive --api-key \"$FAL_KEY\"\n  genmedia models --endpoint_id \u003Ccandidate-id> --json   # confirm it exists\n  genmedia schema  \u003Ccandidate-id> --json                # compare input params\n  genmedia pricing \u003Ccandidate-id> --json                # cost delta\n  ```\n- **fal-ai MCP** if available in this session: `search_models`,\n  `get_model_schema`, `get_pricing` give the same data.\n- **Fallback (zero-auth):** the OpenAPI spec — a 200 means the endpoint is real:\n  `curl -s -o \u002Fdev\u002Fnull -w \"%{http_code}\" \"https:\u002F\u002Ffal.ai\u002Fapi\u002Fopenapi\u002Fqueue\u002Fopenapi.json?endpoint_id=\u003Cid>\"`,\n  and `https:\u002F\u002Ffal.ai\u002Fmodels\u002F\u003Cpath>\u002Fllms.txt` for param specs (see CLAUDE.md).\n\n**text models:** confirm the candidate id resolves on\n`https:\u002F\u002Fopenrouter.ai\u002Fapi\u002Fv1\u002Fmodels` and keeps the same tier (don't turn a\n`-mini` into a non-mini, or a `pro` into `flash`).\n\n**BytePlus Ark ids:** verify against the live ModelArk docs\n(`docs.byteplus.com\u002Fen\u002Fdocs\u002FModelArk\u002F…`) plus the installed adapter's\n`model-meta` (its JSDoc records doc-vs-reality findings, e.g. per-model\nresolution tiers). Ark ids end in a `yymmdd` snapshot date; tier suffixes\n(`-fast`, `-mini`, `-lite`, `-pro`) are different products, not successors.\nTwo Ark-specific gotchas: a new model answers **404 ModelNotOpen until\nactivated in the Ark console** (verify at request time, not startup), and\nresolution tiers are per-model AND per-tier priced — the adapter's model-meta\nand BytePlus's own pricing page have disagreed about which tiers exist (e.g.\nSeedance 2.5's 1080p), so check the official rate table\n(`docs.byteplus.com\u002Fen\u002Fdocs\u002FModelArk\u002F1544106`) for the tier the request\nbuilders ask for, and confirm the rate-card entry matches THAT tier's price.\n\n**Reject a candidate when** it is a different tier\u002Fvariant (fast\u002Flite\u002Fstandard\nvs pro, lora\u002Ftrainer\u002Fedit gear), a different modality, a preview\u002Fexperimental\nbuild replacing a stable one, or its schema\u002Fpricing changed so much it needs\nproduct judgement. When unsure, skip it and note it in the PR body rather than\nguessing.\n\n## 3. Apply the bump (one PR per upgrade)\n\nWork one upgrade at a time so each PR is reviewable and revertible.\n\n**Idempotency — check first:** `gh pr list --state open --search \"in:title model\"`.\nIf an open PR or branch already covers this exact bump, skip it. Branch name:\n`auto\u002Fmodel-update-\u003Cregistry-key>-\u003Cnew-version>` (e.g.\n`auto\u002Fmodel-update-minimax_hailuo_02-2.3`).\n\n**Golden rule: change the `id` (and metadata), never the registry KEY.** Keys\nare persisted in the DB (a team's selected model) and referenced across selectors\nand schemas — renaming one is a breaking migration, out of scope here.\n\nPer class, edit and follow through:\n\n- **Text** (`models.config.ts`): update `id`, `name`, `description`,\n  `contextWindow`. If you bump `DEFAULT_ANALYSIS_MODEL`'s model, the constant\n  references a key, so it's unaffected.\n  **Catalog-lag bridge:** the `@tanstack\u002Fai-openrouter` adapter ships a codegen\n  snapshot of OpenRouter's model list that lags new releases, so a freshly\n  shipped id may not be in its typed union yet. When a text-model bump adopts an\n  id the installed catalog lacks, `bun typecheck` fails at the `createAdapter`\n  call sites — add a `createModel` entry for the id to `CATALOG_LAG_MODELS`\n  (`src\u002Fmodels\u002Fserver\u002Fcreate-adapter.ts`) with the correct `input` modalities, plus\n  `features: ['reasoning', 'structured_outputs']` when the model supports them.\n  (The reverse — pruning a bridged id once the adapter package catches up — is\n  handled by Dependabot's package bump, guided by `catalog-lag.test.ts`, not by\n  this routine.)\n- **Image** (`models.ts` `IMAGE_MODELS`): update `id`, `name`, `description`,\n  `maxPromptLength`. If the model has an `EDIT_ENDPOINTS` entry, update that\n  endpoint id too. Confirm the new endpoint still supports the edit\u002Freference\n  flow if it had one.\n- **Video \u002F motion** (`models.ts` `IMAGE_TO_VIDEO_MODELS`): update `id` + meta,\n  then regenerate the schemas — **`bun motion:codegen`** (writes\n  `src\u002Fmotion\u002Fserver\u002Fgenerated\u002F**` and `src\u002Fmotion\u002Fserver\u002Fendpoint-map.ts`). Never hand-write motion\n  schemas. Re-check `maxPromptLength` against the new schema.\n- **Audio** (`models.ts` `AUDIO_MODELS`): update `id` + `capabilities`\n  (durations, formats) from the new schema.\n- **BytePlus-routed models** (a `byteplusId` on the entry): bump the\n  `byteplusId` in the same PR as the fal id — never one without the other.\n  Then follow through:\n  - `src\u002Fbilling\u002Fbyteplus-pricing.ts` — the rate card is keyed by Ark model id,\n    so the old id's entry must be replaced with the new id at the new model's\n    advertised rate (BytePlus publishes no pricing API; read the pricing page,\n    re-date the header comment, and note the rate is advertised-not-verified).\n    A missed rename means Ark generations bill $0 — exactly the #1069 failure\n    mode the card exists to prevent.\n  - `src\u002Fbilling\u002Ffal-cost.ts` `ENDPOINT_STRATEGY` — if the old fal endpoint ids\n    appear there (token-billed Seedance endpoints do), rename them too.\n  - The request builders (`build-byteplus-video-request.ts` \u002F\n    `build-byteplus-image-request.ts`) read the id from the registry, but\n    re-check their baked-in assumptions (resolution tier, watermark default,\n    frame-vs-reference roles) against the new model's docs.\n  - e2e fixtures live under a fal-endpoint-keyed dir and each fixture's\n    `match.model` is the fal endpoint id, which aimock matches on — migrate\n    them (rename dir + edit `match.model`) rather than re-recording when the\n    prompt is unchanged.\n- **fal pricing:** model ids are pricing keys in `model_pricing` (D1, see `src\u002Fbilling\u002Fserver\u002Ffal-pricing-live.ts`)\n  (auto-generated). After any fal id change run **`bun scripts\u002Fupdate-fal-pricing.ts`**\n  (needs `FAL_KEY`). If it can't run, add the new id's pricing manually via the\n  override path documented in that script and flag it in the PR.\n\n## 4. Quality gates (must pass before opening the PR)\n\n```bash\nbun typecheck\nbun lint\nbun run test src\u002Fmodels src\u002Fmotion   # registry + motion suites\nbun run test src\u002Fbilling             # pricing\u002Fcost if fal pricing changed\n```\n\nFix anything that breaks. If a bump cascades into non-trivial changes (schema\nshape changed, pricing model differs, tests need rework), stop and open an\n**issue** describing it instead of forcing a half-working PR.\n\n## 5. Open the PR\n\n```bash\ngit checkout -b auto\u002Fmodel-update-\u003Ckey>-\u003Cversion>\ngit commit -am \"chore(models): bump \u003Cname> \u003Cold> → \u003Cnew>\"\ngh pr create --title \"chore(models): bump \u003Cname> \u003Cold> → \u003Cnew>\" --body \"\u003Cbody>\"\n```\n\nPR body must include, per change: registry key, old id → new id, why it's a\ngenuine successor, links (fal model page \u002F OpenRouter), pricing delta, and\nwhich generated files \u002F pricing were regenerated. End with: \"🤖 Opened by the\ndaily model-freshness routine (#792). Verify pricing & quality before merge.\"\n\nLeave PRs as **draft** only if a quality gate is amber (e.g. pricing couldn't be\nauto-fetched); otherwise ready-for-review. Do not merge — a human reviews.\n",{"name":5,"description":7},"main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","06-前端与设计\u002Fupdate-model-versions",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill update-model-versions",{"tool":38,"command":39},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill update-model-versions",{"tool":41,"command":42},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill update-model-versions","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fupdate-model-versions\u002FSKILL.md",true,[],[47,54,59,66,73,80],{"skillNo":48,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":19},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],{"skillNo":55,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":19},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]