[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71GX89APRV34VCV6RP79BX4":3,"skill-same:01M71GX89APRV34VCV6RP79BX4":64},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":43,"commitSha":44,"skillDirectory":45,"forksCount":46,"repositoryHtmlUrl":47,"repositoryHomepageUrl":6,"ownerHtmlUrl":48,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":49,"install":50,"resources":63},"01M71GX89APRV34VCV6RP79BX4","plasm-catalog-adversarial-verify",null,"Programming-agent adversarial prove loop for Plasm catalogs (AppWorld-first). Certifies teaching + dry\u002Flive Plasm and mandatory semantic auto-seed (intent-only plasm_context). Writes evidence under apis\u002F...\u002Fe2e\u002F. Use when verifying or hardening a catalog outside CUGA\u002Feval ladders.","PlasmTools\u002Fplasm-core","PlasmTools","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F281353390?v=4","01M71GWJFJFZQRD4NNTXY58ZQ8","01M71GWJG8F2ADWJDZCD57ZH94",9,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-adversarial-verify\u002FSKILL.md",[],[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37],"catalog","adversarial","verify","adversarial verify","plasm catalog adversarial verify","intent","first","dry","programming","teaching","when","loop","use","prove","auto","e2e","only","eval","for","evidence","2026-11-10 10:12:30",false,"---\nname: plasm-catalog-adversarial-verify\ndescription: Programming-agent adversarial prove loop for Plasm catalogs (AppWorld-first). Certifies teaching + dry\u002Flive Plasm and mandatory semantic auto-seed (intent-only plasm_context). Writes evidence under apis\u002F...\u002Fe2e\u002F. Use when verifying or hardening a catalog outside CUGA\u002Feval ladders.\n---\n\n# Plasm catalog adversarial verify\n\nProgramming agent certifies **catalog machine-spirits** (CGS\u002FCML + teaching + discovery seating). This is **outside** CUGA \u002F ladder harnesses — do not treat ladder pass rates as catalog certification.\n\n## Mandatory tiers (full cutover)\n\nComplete **all** tiers. Do not stop after schema validate or seeded HTTP execute.\n\n### 1. Schema \u002F cardinality\n\n- `plasm-cgs schema validate` on the catalog (hard-fail dual-wire pagination, missing blocks).\n- List cardinality: ≤1 `query` + ≤1 `search` per entity (no twin-query heresy).\n\n### 2. Teaching + dry\u002Flive (seeded or execute open)\n\n- Open an execute \u002F MCP session that seats the entities under test.\n- Audit the teaching card: auth gates, shelf discriminants (e.g. LikedSong vs Song), critical mutators\u002Fpolarities expressible.\n- Dry (`plasm` plan \u002F `?mode=plan`) then live a minimal program against Hermit or the real backend.\n\n### 3. Semantic auto-seed (mandatory)\n\nProduct path — **not optional**, not a harness shim:\n\n1. Start `plasm-mcp` (or hosted equivalent) with catalogs packed and **`PLASM_DISCOVERY_SEMANTIC_AUTO_SEED=1`** plus **`OPENROUTER_API_KEY`** (same env family as CUGA treatment \u002F intent-discovery docs).\n2. Confirm auto-seed host shape: `tools\u002Flist` exposes `plasm_context` \u002F `plasm` \u002F `plasm_run` and **omits** `discover_capabilities`; `plasm_context` **`session_mode: \"new\"` rejects `seeds`**.\n3. Call **`plasm_context`** with **`session_mode: \"new\"` + `intent` only** (no seeds) for representative intents mined from `eval\u002Fcases.yaml` and ladder-relevant scenarios.\n4. Assert:\n   - Teaching card seats the right entities (auth co-seed, shelf discriminants, federated supervisor passwords when AppWorld).\n   - No `hard_miss` \u002F unjustified `clarify` \u002F brand-lock routing errors.\n   - Critical mutators\u002Fpolarities remain expressible on the card.\n5. Dry (then live when backend is up) a **minimal program from that card** using taught `e#` forms — not hand-invented seeds.\n\n**Fix CGS\u002Fdiscovery when seating is wrong** (`discovery.names`, descriptions, `seed_class`, `co_seed_with`, shelf copy). Do **not** paper over with harness hacks, forced seeds, or prompt-only patches.\n\nBrand-lock \u002F named-catalog matching must be **whole-token** (short ids like `phone` must not fire inside `headphones`).\n\n### 4. Evidence\n\nPer API, write\u002Fupdate `apis\u002F\u003Capi>\u002Fe2e\u002Fadversarial-verify-\u003Cdate>.md` including a **Semantic seed** section:\n\n- pass\u002Ffail\n- intent used\n- what was seated (`e#` → entity)\n- dry verdict \u002F program used\n\nRoll up multi-API work in `apis\u002F\u003Cfamily>\u002Fe2e\u002Fadversarial-roll-\u003Cdate>.md` with a **Semantic seed** column.\n\n## Anti-patterns\n\n- Certifying from manual `{api, entity}` seeds only.\n- Skipping semantic auto-seed because “schema validate passed”.\n- Teaching non-executable sugar (`[…]` legacy projections) without noting runtime rejection.\n- Dual-path shims “for later” — full cutover.\n\n## Related\n\n- Authoring core: [plasm-authoring](..\u002Fplasm-authoring\u002FSKILL.md)\n- Intent discovery \u002F auto-seed: repo `docs\u002Fintent-discovery.md`\n- E2E transport loop: [plasm-catalog-e2e-test](..\u002Fplasm-catalog-e2e-test\u002FSKILL.md)\n","# Plasm catalog adversarial verify\n\nProgramming agent certifies **catalog machine-spirits** (CGS\u002FCML + teaching + discovery seating). This is **outside** CUGA \u002F ladder harnesses — do not treat ladder pass rates as catalog certification.\n\n## Mandatory tiers (full cutover)\n\nComplete **all** tiers. Do not stop after schema validate or seeded HTTP execute.\n\n### 1. Schema \u002F cardinality\n\n- `plasm-cgs schema validate` on the catalog (hard-fail dual-wire pagination, missing blocks).\n- List cardinality: ≤1 `query` + ≤1 `search` per entity (no twin-query heresy).\n\n### 2. Teaching + dry\u002Flive (seeded or execute open)\n\n- Open an execute \u002F MCP session that seats the entities under test.\n- Audit the teaching card: auth gates, shelf discriminants (e.g. LikedSong vs Song), critical mutators\u002Fpolarities expressible.\n- Dry (`plasm` plan \u002F `?mode=plan`) then live a minimal program against Hermit or the real backend.\n\n### 3. Semantic auto-seed (mandatory)\n\nProduct path — **not optional**, not a harness shim:\n\n1. Start `plasm-mcp` (or hosted equivalent) with catalogs packed and **`PLASM_DISCOVERY_SEMANTIC_AUTO_SEED=1`** plus **`OPENROUTER_API_KEY`** (same env family as CUGA treatment \u002F intent-discovery docs).\n2. Confirm auto-seed host shape: `tools\u002Flist` exposes `plasm_context` \u002F `plasm` \u002F `plasm_run` and **omits** `discover_capabilities`; `plasm_context` **`session_mode: \"new\"` rejects `seeds`**.\n3. Call **`plasm_context`** with **`session_mode: \"new\"` + `intent` only** (no seeds) for representative intents mined from `eval\u002Fcases.yaml` and ladder-relevant scenarios.\n4. Assert:\n   - Teaching card seats the right entities (auth co-seed, shelf discriminants, federated supervisor passwords when AppWorld).\n   - No `hard_miss` \u002F unjustified `clarify` \u002F brand-lock routing errors.\n   - Critical mutators\u002Fpolarities remain expressible on the card.\n5. Dry (then live when backend is up) a **minimal program from that card** using taught `e#` forms — not hand-invented seeds.\n\n**Fix CGS\u002Fdiscovery when seating is wrong** (`discovery.names`, descriptions, `seed_class`, `co_seed_with`, shelf copy). Do **not** paper over with harness hacks, forced seeds, or prompt-only patches.\n\nBrand-lock \u002F named-catalog matching must be **whole-token** (short ids like `phone` must not fire inside `headphones`).\n\n### 4. Evidence\n\nPer API, write\u002Fupdate `apis\u002F\u003Capi>\u002Fe2e\u002Fadversarial-verify-\u003Cdate>.md` including a **Semantic seed** section:\n\n- pass\u002Ffail\n- intent used\n- what was seated (`e#` → entity)\n- dry verdict \u002F program used\n\nRoll up multi-API work in `apis\u002F\u003Cfamily>\u002Fe2e\u002Fadversarial-roll-\u003Cdate>.md` with a **Semantic seed** column.\n\n## Anti-patterns\n\n- Certifying from manual `{api, entity}` seeds only.\n- Skipping semantic auto-seed because “schema validate passed”.\n- Teaching non-executable sugar (`[…]` legacy projections) without noting runtime rejection.\n- Dual-path shims “for later” — full cutover.\n\n## Related\n\n- Authoring core: [plasm-authoring](..\u002Fplasm-authoring\u002FSKILL.md)\n- Intent discovery \u002F auto-seed: repo `docs\u002Fintent-discovery.md`\n- E2E transport loop: [plasm-catalog-e2e-test](..\u002Fplasm-catalog-e2e-test\u002FSKILL.md)\n",{"name":5,"description":7},"main","8f036e9443513eb615ca40218052395b4f489eba","skills\u002Fplasm-catalog-adversarial-verify",3,"https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core","https:\u002F\u002Fgithub.com\u002FPlasmTools","2026-11-12 03:32:20",{"source":8,"skillName":5,"commands":51,"prompt":61,"zipAvailable":62},[52,55,58],{"tool":53,"command":54},"npx","npx skills add PlasmTools\u002Fplasm-core --skill plasm-catalog-adversarial-verify",{"tool":56,"command":57},"bunx","bunx skills add PlasmTools\u002Fplasm-core --skill plasm-catalog-adversarial-verify",{"tool":59,"command":60},"pnpm","pnpm dlx skills add PlasmTools\u002Fplasm-core --skill plasm-catalog-adversarial-verify","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-adversarial-verify\u002FSKILL.md",true,[],[65,92,99,107,114,122],{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":76,"publishedAt":38},"01M71GXN906T8EMB9G6PZ5XA38","plasm-forge","OSS entry skill for Plasm CGS \u002F CML catalog authoring — points at the canonical skill suite under `skills\u002F` in plasm-core. Use when agents need the authoring playbook; pair with `.cursor\u002Fagents\u002Fplasm-forge.md` (Cursor) for autonomous catalog work.","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-forge\u002FSKILL.md",[71],{"categoryNo":72,"categoryName":73,"taxonomyType":74,"taxonomyTypeDesc":75},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[77,78,79,80,81,82,83,84,85,86,87,88,89,28,30,90,18,36,91],"plasm forge","forge","pair","oss","canonical","suite","authoring","playbook","entry","at","cursor","work","with","autonomous","core",{"skillNo":93,"name":94,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":38},"01M71GXKKSBRV99VB8JHXTSZ5V","plasm-catalog-score","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-score\u002FSKILL.md",[],[98,18],"plasm catalog score",{"skillNo":100,"name":101,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":38},"01M71GXHASQBWNG6R5CTDJEZZW","plasm-catalog-retro","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-retro\u002FSKILL.md",[],[105,106,18],"retro","plasm catalog retro",{"skillNo":108,"name":109,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":110,"categories":111,"keywords":112,"publishedAt":38},"01M71GXFM75YTJDHM24BY9ERHA","plasm-catalog-reprint","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-reprint\u002FSKILL.md",[],[113,18],"plasm catalog reprint",{"skillNo":115,"name":116,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":117,"categories":118,"keywords":119,"publishedAt":38},"01M71GXDC86B9XJHTZ385K4F5E","plasm-catalog-polish","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-polish\u002FSKILL.md",[],[120,121,18],"polish","plasm catalog polish",{"skillNo":123,"name":124,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":125,"categories":126,"keywords":130,"publishedAt":38},"01M71GXBJV6YV7H86Y4T9CBXK3","plasm-catalog-e2e-test","https:\u002F\u002Fgithub.com\u002FPlasmTools\u002Fplasm-core\u002Fblob\u002Fmain\u002Fskills\u002Fplasm-catalog-e2e-test\u002FSKILL.md",[127],{"categoryNo":128,"categoryName":129,"taxonomyType":74,"taxonomyTypeDesc":75},"01JBSEEDCAT000000000000008","测试与质量",[18,131,132,133,33],"e2e test","plasm catalog e2e test","test"]