[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M759WGDM3KZZTQJX8VTMJAC8":3,"skill-same:01M759WGDM3KZZTQJX8VTMJAC8":54},{"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":31,"commitSha":32,"skillDirectory":33,"forksCount":34,"repositoryHtmlUrl":35,"repositoryHomepageUrl":6,"ownerHtmlUrl":36,"ownerBlogUrl":37,"ownerDescription":38,"updatedAt":39,"install":40,"resources":53},"01M759WGDM3KZZTQJX8VTMJAC8","sota-review",null,"Survey the outside world before designing — TWO lenses in one skill. Papers (vault\u002Frepo grep first, then the arXiv API over HTTPS, never search-engine scraping) and codebases (the GitHub API against the watched-repo registry's canonical repos, never a local fork checkout). Cited, provenance-stamped briefs. Use before specs, decompositions, and hypothesis rounds.","harmoniqs\u002Famicode","harmoniqs","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F199011497?v=4","01M759PQCJ9Y093EC5HC3344DW","Harmoniqs","01M759PQD5S85W68FZKV3XD1X0",7,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Fskills\u002Fsota-review\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: sota-review\ndescription: Survey the outside world before designing — TWO lenses in one skill. Papers (vault\u002Frepo grep first, then the arXiv API over HTTPS, never search-engine scraping) and codebases (the GitHub API against the watched-repo registry's canonical repos, never a local fork checkout). Cited, provenance-stamped briefs. Use before specs, decompositions, and hypothesis rounds.\nagents: [researcher, librarian, dreamer]\nsurface: public\nsource: amicode\nrevision: 1\n---\n\n# SOTA review — the loops' external currency\n\nOne survey, two lenses, one discipline: **read-only toward the world,\nappend-only toward the vault**. The survey seeds hypotheses and design\ndecisions; it never blocks the loop (a survey that cannot run records a\nNAMED outcome and the loop proceeds), and its matches **stage before they\ncount**: the daily digest and the SOTA watcher append `stage` lines to\nper-campaign SIDECAR staging streams beside the session ledgers, and nothing\ncounts as currency until the PI's accept stamp lands. (The session ledgers\nhere are the campaign session ledgers — running ledgers in the personal\nvault's `sessions\u002F` directory — never the retired transcript substrate.)\nThis skill's lenses fetch and report — they never append; the staged routing\nis its own layer.\n\n## Usage\n\n`\u002Fsota-review` — survey papers and watched codebases for the topic at hand.\n`\u002Fsota-review papers \u003Cterms>` — the papers lens only.\n`\u002Fsota-review codebase` — the codebase lens only (a watched-repo round).\n\n## Lens 1 — papers\n\nThe recipe's rules, in order. **Step 1 always precedes step 2** — niche\nqueries are usually already answered in the vault, the demos, or the\npackage docs.\n\n1. **Vault\u002Frepo grep FIRST.** Grep the vault mounts and the repo checkouts\n   before touching the network:\n   `grep -rn -i \"\u003Cterm>\" \u003Cvault mounts> \u003Crepo checkouts>`\n   Vendors, collaborators, and papers-we-cite are typically already in\n   notes, reading cards, or package docs — a hit here is faster and better\n   than any web result.\n2. **The arXiv API over HTTPS for papers** — scriptable, reliable, never\n   scraping. On-demand queries go through the fleet-wide serialized queue:\n   `amico sota papers --query \"\u003Cterms>\" [--top N]`\n   GOTCHA (mechanical, not advisory): the `http:\u002F\u002F` scheme on\n   export.arxiv.org **silently hangs** — always `https:\u002F\u002F`. The fetch seam\n   refuses `http:\u002F\u002F` by name; do not work around it.\n3. **Direct domain or Wikipedia fetch** for companies and tools: fetch the\n   plausible domain or `en.wikipedia.org\u002Fwiki\u002F\u003CName>`. A 404 is signal too.\n4. **Ask the user** when 1–3 fail — a one-line question beats ten minutes\n   of scraper thrash.\n\n**Never search-engine scraping.** Bing or Google HTML endpoints are SEO\ngarbage for niche technical queries or bot-blocked outright — never fetch\nthem, never parse them. The sanctioned paths are the two above.\n\n### The queue discipline (one fetcher, one query queue)\n\nAll live arXiv traffic rides the fetch cache plus ONE serialized query queue\n— a lock-file queue at the fleet-shared sota root (the personal vault\nmount's `amicode\u002Fsota` directory; `$AMICO_SOTA_ROOT` overrides). A per-host\nlock serializes nothing: the fleet is the concurrency, so the queue lives\nat the shared path. In practice:\n\n- If the cache holds the query, you get `via: cache` — no network ran.\n- If another fetcher holds the queue, the wait is bounded; on timeout you\n  get the **named outcome** `queue-timeout` with the disclosed alternative\n  (read the fetch cache, or record the explicit waiver). Never retry in a\n  tight loop; never block the loop on the wire.\n- **Known limit** (named, not hidden): the lock is atomic per host —\n  vault-sync propagation latency makes fleet-wide mutual exclusion\n  best-effort, and the under-lock cache re-check bounds the damage to a\n  duplicate fetch, never a correctness break.\n\n## Lens 2 — codebases\n\nWatch the repos that matter via the GitHub API **against canonical repos\n(organizer\u002Fname), never a local fork checkout** — a local checkout sees\nyour drift, not the world's:\n\n`amico sota codebase [--repo owner\u002Fname]`\n\nThe substrate is the **watched-repo registry** — validator-checked TOML\ndata living in the same fleet-shared sota root (`watched-repos.toml`),\nseeded at first read with the shipped set: the canonical agent harnesses,\nthe Julia optimal-control stack adjacent to our authoring map, the\nQEC\u002FqLDPC challenge repos, and the quantum SDK release tracks.\n\n- **Adding a repo is a data edit, never code**: append a `[[repos]]` entry\n  (`repo`, `why_watched`, `domains`, `fetch_surface`, `match_keywords`).\n  A malformed registry fails LOUDLY — fix the TOML, never the validator.\n- Each round stamps `last_success` and accrues `consecutive_failures` per\n  entry; after the registry's threshold (default 7) of consecutive failures\n  the brief flags the entry for **human retire-or-confirm** — confirm the\n  watch or retire it; the registry edit is yours, never the machinery's.\n- Events surface only when they match the entry's keywords, and the brief\n  names which keywords matched.\n\n## Briefs — the PI register\n\nEvery brief this skill emits is written for a research PI: **concise,\nhuman-readable, details-informed**, every claim **cited** (the paper's abs\nURL, the release\u002Fissue URL), and every render **provenance-stamped** — the\nquery, the source, the fetch time, and the path (`via: cache` or\n`via: fetched`; a cache read never launders as a live fetch). Raw payloads\nstay one level down; the brief ranks, it never dumps.\n\n**Quiet failures are the ones that matter:**\n\n- A successful fetch returning zero entries where the trailing 7-fetch-day\n  mean is nonzero renders **\"scan returned nothing — anomalous\"** — never\n  \"nothing new\". An empty scan against history is a signal, not a rest day.\n- A failed fetch is a **named** failure in the brief (which repo, which\n  surface, which error) — never a silent skip.\n\n## Checklist — the recipe's rules (lint-pinned)\n\nRun through this list before publishing any brief:\n\n- [ ] Vault and repo greps ran FIRST, and their hits (if any) are cited in the brief.\n- [ ] Papers came from the arXiv API over **HTTPS** — `https:\u002F\u002Fexport.arxiv.org\u002Fapi\u002Fquery`, never `http:\u002F\u002F` (it hangs), never search-engine HTML (no Bing, no Google scraping).\n- [ ] Every live fetch rode the serialized queue (`amico sota …`); a `queue-timeout` rendered its named outcome and the disclosed alternative — the loop was never blocked.\n- [ ] Codebase events came from the GitHub API against **canonical repos**, never a local fork checkout.\n- [ ] The registry is treated as data: any repo change was a TOML edit; a flagged entry went to the human as retire-or-confirm.\n- [ ] Every claim in the brief is cited and the render carries its provenance stamp (query, source, fetched time, via).\n- [ ] An empty-but-successful scan rendered \"scan returned nothing — anomalous\" if the floor was armed — the brief never says \"nothing new\".\n\n## Honest degradation\n\n- **No sota root \u002F no registry**: the codebase lens bootstraps the registry\n  from the shipped seed on first read; if even that fails, say so and name\n  the error — never pretend a survey ran.\n- **Network unavailable**: the survey records the named outcome and the\n  loop proceeds (currency is a seed, not a stall). The brief renders the\n  failure and its provenance honestly.\n- **The lenses fetch and report; they never append.** Staged routing is its\n  own layer, and its writers are enumerated: the papers digest\n  (`amico papers digest --route`) and the SOTA watcher\n  (`amico sota watcher`) are the match writers; the **accept stamp**\n  (`amico sota accept`, recording the PI's explicit instruction) is the only\n  sanctioned other writer. The **awaiting-the-eye listing**\n  (`amico sota awaiting-the-eye`) renders everything pending — staged\n  matches are never rendered as currency until accepted. No strategy\n  composition renders here either — that layer lives elsewhere; do not\n  simulate its effects.\n","# SOTA review — the loops' external currency\n\nOne survey, two lenses, one discipline: **read-only toward the world,\nappend-only toward the vault**. The survey seeds hypotheses and design\ndecisions; it never blocks the loop (a survey that cannot run records a\nNAMED outcome and the loop proceeds), and its matches **stage before they\ncount**: the daily digest and the SOTA watcher append `stage` lines to\nper-campaign SIDECAR staging streams beside the session ledgers, and nothing\ncounts as currency until the PI's accept stamp lands. (The session ledgers\nhere are the campaign session ledgers — running ledgers in the personal\nvault's `sessions\u002F` directory — never the retired transcript substrate.)\nThis skill's lenses fetch and report — they never append; the staged routing\nis its own layer.\n\n## Usage\n\n`\u002Fsota-review` — survey papers and watched codebases for the topic at hand.\n`\u002Fsota-review papers \u003Cterms>` — the papers lens only.\n`\u002Fsota-review codebase` — the codebase lens only (a watched-repo round).\n\n## Lens 1 — papers\n\nThe recipe's rules, in order. **Step 1 always precedes step 2** — niche\nqueries are usually already answered in the vault, the demos, or the\npackage docs.\n\n1. **Vault\u002Frepo grep FIRST.** Grep the vault mounts and the repo checkouts\n   before touching the network:\n   `grep -rn -i \"\u003Cterm>\" \u003Cvault mounts> \u003Crepo checkouts>`\n   Vendors, collaborators, and papers-we-cite are typically already in\n   notes, reading cards, or package docs — a hit here is faster and better\n   than any web result.\n2. **The arXiv API over HTTPS for papers** — scriptable, reliable, never\n   scraping. On-demand queries go through the fleet-wide serialized queue:\n   `amico sota papers --query \"\u003Cterms>\" [--top N]`\n   GOTCHA (mechanical, not advisory): the `http:\u002F\u002F` scheme on\n   export.arxiv.org **silently hangs** — always `https:\u002F\u002F`. The fetch seam\n   refuses `http:\u002F\u002F` by name; do not work around it.\n3. **Direct domain or Wikipedia fetch** for companies and tools: fetch the\n   plausible domain or `en.wikipedia.org\u002Fwiki\u002F\u003CName>`. A 404 is signal too.\n4. **Ask the user** when 1–3 fail — a one-line question beats ten minutes\n   of scraper thrash.\n\n**Never search-engine scraping.** Bing or Google HTML endpoints are SEO\ngarbage for niche technical queries or bot-blocked outright — never fetch\nthem, never parse them. The sanctioned paths are the two above.\n\n### The queue discipline (one fetcher, one query queue)\n\nAll live arXiv traffic rides the fetch cache plus ONE serialized query queue\n— a lock-file queue at the fleet-shared sota root (the personal vault\nmount's `amicode\u002Fsota` directory; `$AMICO_SOTA_ROOT` overrides). A per-host\nlock serializes nothing: the fleet is the concurrency, so the queue lives\nat the shared path. In practice:\n\n- If the cache holds the query, you get `via: cache` — no network ran.\n- If another fetcher holds the queue, the wait is bounded; on timeout you\n  get the **named outcome** `queue-timeout` with the disclosed alternative\n  (read the fetch cache, or record the explicit waiver). Never retry in a\n  tight loop; never block the loop on the wire.\n- **Known limit** (named, not hidden): the lock is atomic per host —\n  vault-sync propagation latency makes fleet-wide mutual exclusion\n  best-effort, and the under-lock cache re-check bounds the damage to a\n  duplicate fetch, never a correctness break.\n\n## Lens 2 — codebases\n\nWatch the repos that matter via the GitHub API **against canonical repos\n(organizer\u002Fname), never a local fork checkout** — a local checkout sees\nyour drift, not the world's:\n\n`amico sota codebase [--repo owner\u002Fname]`\n\nThe substrate is the **watched-repo registry** — validator-checked TOML\ndata living in the same fleet-shared sota root (`watched-repos.toml`),\nseeded at first read with the shipped set: the canonical agent harnesses,\nthe Julia optimal-control stack adjacent to our authoring map, the\nQEC\u002FqLDPC challenge repos, and the quantum SDK release tracks.\n\n- **Adding a repo is a data edit, never code**: append a `[[repos]]` entry\n  (`repo`, `why_watched`, `domains`, `fetch_surface`, `match_keywords`).\n  A malformed registry fails LOUDLY — fix the TOML, never the validator.\n- Each round stamps `last_success` and accrues `consecutive_failures` per\n  entry; after the registry's threshold (default 7) of consecutive failures\n  the brief flags the entry for **human retire-or-confirm** — confirm the\n  watch or retire it; the registry edit is yours, never the machinery's.\n- Events surface only when they match the entry's keywords, and the brief\n  names which keywords matched.\n\n## Briefs — the PI register\n\nEvery brief this skill emits is written for a research PI: **concise,\nhuman-readable, details-informed**, every claim **cited** (the paper's abs\nURL, the release\u002Fissue URL), and every render **provenance-stamped** — the\nquery, the source, the fetch time, and the path (`via: cache` or\n`via: fetched`; a cache read never launders as a live fetch). Raw payloads\nstay one level down; the brief ranks, it never dumps.\n\n**Quiet failures are the ones that matter:**\n\n- A successful fetch returning zero entries where the trailing 7-fetch-day\n  mean is nonzero renders **\"scan returned nothing — anomalous\"** — never\n  \"nothing new\". An empty scan against history is a signal, not a rest day.\n- A failed fetch is a **named** failure in the brief (which repo, which\n  surface, which error) — never a silent skip.\n\n## Checklist — the recipe's rules (lint-pinned)\n\nRun through this list before publishing any brief:\n\n- [ ] Vault and repo greps ran FIRST, and their hits (if any) are cited in the brief.\n- [ ] Papers came from the arXiv API over **HTTPS** — `https:\u002F\u002Fexport.arxiv.org\u002Fapi\u002Fquery`, never `http:\u002F\u002F` (it hangs), never search-engine HTML (no Bing, no Google scraping).\n- [ ] Every live fetch rode the serialized queue (`amico sota …`); a `queue-timeout` rendered its named outcome and the disclosed alternative — the loop was never blocked.\n- [ ] Codebase events came from the GitHub API against **canonical repos**, never a local fork checkout.\n- [ ] The registry is treated as data: any repo change was a TOML edit; a flagged entry went to the human as retire-or-confirm.\n- [ ] Every claim in the brief is cited and the render carries its provenance stamp (query, source, fetched time, via).\n- [ ] An empty-but-successful scan rendered \"scan returned nothing — anomalous\" if the floor was armed — the brief never says \"nothing new\".\n\n## Honest degradation\n\n- **No sota root \u002F no registry**: the codebase lens bootstraps the registry\n  from the shipped seed on first read; if even that fails, say so and name\n  the error — never pretend a survey ran.\n- **Network unavailable**: the survey records the named outcome and the\n  loop proceeds (currency is a seed, not a stall). The brief renders the\n  failure and its provenance honestly.\n- **The lenses fetch and report; they never append.** Staged routing is its\n  own layer, and its writers are enumerated: the papers digest\n  (`amico papers digest --route`) and the SOTA watcher\n  (`amico sota watcher`) are the match writers; the **accept stamp**\n  (`amico sota accept`, recording the PI's explicit instruction) is the only\n  sanctioned other writer. The **awaiting-the-eye listing**\n  (`amico sota awaiting-the-eye`) renders everything pending — staged\n  matches are never rendered as currency until accepted. No strategy\n  composition renders here either — that layer lives elsewhere; do not\n  simulate its effects.\n",{"name":5,"agents":24,"source":28,"surface":29,"revision":30,"description":7},[25,26,27],"researcher","librarian","dreamer","amicode","public",1,"main","bf84fdd4c1494800695bd5881702b3118dbf98be","packages\u002Fextension\u002Fskills\u002Fsota-review",2,"https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode","https:\u002F\u002Fgithub.com\u002Fharmoniqs","harmoniqs.ai","research tools, for researchers, by reseachers","2026-11-12 14:46:24",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":52},[42,45,48],{"tool":43,"command":44},"npx","npx skills add harmoniqs\u002Famicode --skill sota-review",{"tool":46,"command":47},"bunx","bunx skills add harmoniqs\u002Famicode --skill sota-review",{"tool":49,"command":50},"pnpm","pnpm dlx skills add harmoniqs\u002Famicode --skill sota-review","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Fskills\u002Fsota-review\u002FSKILL.md",true,[],[55,62,69,75,82,89],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":19},"01M759XGHEQRJ1GWZTPRFF9Z1W","malformed","Frontmatter that parses as YAML but is missing the required name field.","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Ftest\u002Ffixtures\u002Fskill-drift\u002Fskills\u002Fmalformed\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M759XDVQ1PYPMT876X8J21TQ","duplicated","Second skill carrying the duplicated frontmatter name.","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Ftest\u002Ffixtures\u002Fskill-drift\u002Fskills\u002Fdup-two\u002FSKILL.md",[],[],{"skillNo":70,"name":64,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":19},"01M759XBZYBHF38JDWSRHAC6E1","First skill carrying the duplicated frontmatter name.","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Ftest\u002Ffixtures\u002Fskill-drift\u002Fskills\u002Fdup-one\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M759XAQBSP29BKTTXR1WF2D7","drifted","A skill whose API claims have drifted from the fixture packages.","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Ftest\u002Ffixtures\u002Fskill-drift\u002Fskills\u002Fdrifted\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M759X8CNBPVK98C6M45D0QMB","clean","A clean skill whose every extracted claim resolves against the fixture packages.","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Ftest\u002Ffixtures\u002Fskill-drift\u002Fskills\u002Fclean\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":19},"01M759X73H6N2G0VDH8CJ5TF11","broken-paths","A skill with a broken relative link and a broken cited package path.","https:\u002F\u002Fgithub.com\u002Fharmoniqs\u002Famicode\u002Fblob\u002Fmain\u002Fpackages\u002Fextension\u002Ftest\u002Ffixtures\u002Fskill-drift\u002Fskills\u002Fbroken-paths\u002FSKILL.md",[],[]]