[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-anthropic-commerce-prompt-caching":3,"mdc--9fldbh-key":37,"related-repo-anthropic-commerce-prompt-caching":1036,"related-org-anthropic-commerce-prompt-caching":1142},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":12,"stars":26,"repoUrl":27,"updatedAt":28,"license":29,"forks":30,"topics":31,"repo":32,"sourceUrl":35,"mdContent":36},"commerce-prompt-caching","configure commerce agent prompt caching","The reference agents' cache-stable request assembly, covering the static system and per-request context split, the fixed tool list, the rolling conversation breakpoint, which config fields are prompt bytes, and verification. Load when writing or reviewing a commerce agent's prompt assembly, when cache reads are zero, or when a turn's latency or cost is the question.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},"anthropic","Anthropic","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fanthropic.png","anthropics",[13,17,20,23],{"name":14,"slug":15,"type":16},"Performance","performance","tag",{"name":18,"slug":19,"type":16},"Prompt Engineering","prompt-engineering",{"name":21,"slug":22,"type":16},"E-commerce","e-commerce",{"name":24,"slug":25,"type":16},"Agents","agents",1735,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcommerce-agents","2026-09-04T08:00:39.50364",null,277,[],{"repoUrl":27,"stars":26,"forks":30,"topics":33,"description":34},[],"Reference blueprint for building shopping and merchant agents with Claude. Examples in retail, commerce, telecom, and entertainment included.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fcommerce-agents\u002Ftree\u002FHEAD\u002Fplugins\u002Fcommerce-builder\u002Fskills\u002Fcommerce-prompt-caching","---\nname: commerce-prompt-caching\ndescription: The reference agents' cache-stable request assembly, covering the static system and per-request context split, the fixed tool list, the rolling conversation breakpoint, which config fields are prompt bytes, and verification. Load when writing or reviewing a commerce agent's prompt assembly, when cache reads are zero, or when a turn's latency or cost is the question.\n---\n\n# Cache-stable request assembly\n\n`commerce_common\u002F` is `commerce-common\u002Fcommerce_common\u002F`, `shopping_agent\u002F` is `shopping-agent\u002Fcore\u002Fshopping_agent\u002F`, and\n`merchant_agent\u002F` is `merchant-agent\u002Fcore\u002Fmerchant_agent\u002F`.\n\n## The three breakpoints\n\nA request's cacheable prefix runs `tools`, then `system`, then the messages; a breakpoint ends a span the next call reads\nback. The reference agents place three, all from `commerce_common\u002Fprompt_assembly.py`:\n\n1. **The last tool**: `with_tool_cache_control(tools)`.\n2. **The static system text**: `build_system_blocks(static, context)` marks it and appends the context block unmarked.\n3. **The newest persisted message**: `build_request_messages(messages, rolling_breakpoint=...)` marks the last persisted\n   block on the outgoing request only; the next call then reads the earlier rounds, search payloads included, from cache.\n\nEverything per request goes in the context block, which the role's `build_dynamic_context(...)` renders once per turn.\nIn the static block or the tool list, those bytes would break the system or tool span on every request; in the context block they cost\none re-read of the conversation on a turn whose cart, page, or facts moved, and the rolling marker holds from that turn's\nnext round. The clock renders to the hour (`context_clock`), so a new minute moves nothing.\n\n## The split as implemented\n\n- `build_static_system(config, skills)` in each role's `prompt.py` renders identity, the most-turns rules, the fence\n  notice, and the skill index from the config and the installed skills: the same bytes for the life of a deployment.\n- `build_dynamic_context(...)` in the same module renders the per-request material inside the role's data fence: shopping\n  takes preferences, memory facts, the cart, the page, the account block, and the time; merchant takes the store context,\n  memory facts, and the time. Backend context blocks have their own size cap.\n- `ShoppingAgent.__init__` and `MerchantAgent.__init__` (each role's `runtime-messages-api` `orchestrator.py`) build\n  `_static_system` and `_tools` once per process; `stream_turn` renders the context block per turn and calls\n  `build_request_messages` per model call.\n- `build_tools` in each role's `tools\u002Fregistry.py` emits the built-ins in a fixed order (the merchant builder adds `run_analysis` after them when enabled), the `load_skill` enum sorted,\n  then extensions in the order given, then `web_search` when enabled; every registered tool ships on every request, and\n  the executor decides on arrival whether a call can be served.\n- The Agent SDK runtimes pass the same static text (plus `SKILL_TOOL_ADAPTER`) and the same contracts, and the SDK caches\n  them; the hosted manifests carry the text as `system.md`, which `scripts\u002Fcheck.py` compares with the builder.\n\n## When the rolling marker is skipped\n\nTwo rounds carry no marker. A bare first call (one message): a one-shot session would pay the write without a read, and\nthe second call's marker covers the first message anyway. A round whose `tool_choice` is other than `auto` (the\ngrounding-forced first iteration and the forced-text last one): `tool_choice` keys the messages span, so an entry written\nunder a forced round is unreadable by the auto rounds after it; the system and tool spans still hit. Both live in\n`build_request_messages` and the loops' `rolling_breakpoint=` argument; `rolling_conversation_cache` turns the marker off\nfor debugging.\n\n## When the conversation is compacted\n\nWhen a turn's last call was given `compact_history_above_tokens` or more (its usage says; the default is the platform's\nown tool-result-clearing default, a tenth of the window), the turn ends with `compact_history` in\n`commerce_common\u002Fturn.py`, which replaces the oldest tool results in the stored conversation with a one-line marker\nuntil the conversation is half its previous size. The next turn's first round rewrites the messages span once and later\nrounds read the shorter one; `turn_complete.results_cleared` tells a host that appends its transcript to rewrite it. The\nsystem and tool spans, the messages, and the write gates are unaffected; provenance lives on the session state.\n\n## Config fields that are prompt bytes\n\nThe fields marked `(prompt)` in `commerce_common\u002Fconfig.py` and the role configs, plus `enable_analysis` and\n`max_items_per_change`, which the merchant tool builder reads, render into the static text or the tool list; changing one is a redeploy and a miss on the next request:\n\n| Config | Fields |\n|---|---|\n| `BaseAgentConfig` (`commerce_common\u002Fconfig.py`) | `brand_name`, `assistant_name`, `brand_voice`, `enable_web_search` |\n| `ShoppingAgentConfig` (`shopping_agent\u002Fconfig.py`) | `domain_search_notes`, `enable_disclosures`, and the system switches `enable_cart`, `enable_orders`, `enable_policies`, `enable_fulfillment` |\n| `MerchantAgentConfig` (`merchant_agent\u002Fconfig.py`) | `enable_analysis`, `require_host_approval`, `approval_surface`, `stage_shows_preview`, and the system switches `enable_listing_edits`, `enable_inventory`, `enable_pricing`, `enable_campaigns`; `max_items_per_change` and `max_search_results` set `maxItems` and `maximum` on tool schemas |\n\nSkills, presentation extensions, and delegates are prompt bytes too. Gate lexicons, guardrail limits, memory settings,\nthe latency knobs (`eager_tool_dispatch`, `rolling_conversation_cache`, `eager_partial_frames`,\n`close_on_presentation`), cart caps, and `compact_history_above_tokens` are not; `tests\u002Ftest_role_registries.py`\nasserts it for each.\n\n## What breaks a hit\n\n- Anything per request in the static block or the tool list: a name, a cart count, a page, a clock, a request id. The\n  context block takes the first four; a request id belongs nowhere.\n- A set or dict iterated into prompt text or a schema without sorting.\n- `tools[]` membership decided per request (a flag read per request instead of once at construction).\n- Rebuilding the static text or the tool list inside the turn instead of in the constructor.\n- Persisting the rolling marker into the stored conversation, which then gains one marker per turn.\n- Toggling a `(prompt)` field, a skill, or an extension on a running deployment; each is a redeploy.\n- A prompt variant chosen per request instead of per deployment.\n- A prefix under the model's minimum cacheable length, which writes nothing.\n- A forced `tool_choice` (grounding first, `none` last) misses the messages span only; the loops skip the marker there.\n\n## How to verify\n\n- `turn_complete` carries `usage` (`usage_totals` in `commerce_common\u002Fturn.py`, the turn's calls summed) and\n  `elapsed_ms`; every call also logs one line with the same counters and its own time (`log_model_call`, on\n  `shopping_agent_runtime.orchestrator` and `merchant_agent_runtime.orchestrator`). Zero cache reads on the second turn\n  of a conversation means the prefix changed. The counters worth charting per config version are cache reads as a\n  share of input, `elapsed_ms`, rounds per turn, and blocked `tool_result` events by gate.\n- `tests\u002Ftest_role_registries.py` builds each role's prompt and tools twice and compares bytes, checks the cache marks,\n  and checks that non-prompt settings change nothing; `commerce-common\u002Ftests\u002Ftest_prompt_assembly.py` covers the block\n  builders, the context clock, the rolling marker, and the skip cases; `tests\u002Ftest_turn_loop.py` pins both loops to the\n  same blocks, marker, and clean history. A deployment's tests carry the same three checks.\n- In a deployed environment, run one three-turn conversation and read the second and third turns' usage; a proxy or retry\n  layer that rewrites requests shows up here and nowhere else.\n",{"data":38,"body":39},{"name":4,"description":6},{"type":40,"children":41},"root",[42,51,102,109,138,195,216,222,392,398,449,455,492,498,535,787,836,842,918,924],{"type":43,"tag":44,"props":45,"children":47},"element","h1",{"id":46},"cache-stable-request-assembly",[48],{"type":49,"value":50},"text","Cache-stable request assembly",{"type":43,"tag":52,"props":53,"children":54},"p",{},[55,62,64,70,72,78,79,85,87,93,94,100],{"type":43,"tag":56,"props":57,"children":59},"code",{"className":58},[],[60],{"type":49,"value":61},"commerce_common\u002F",{"type":49,"value":63}," is ",{"type":43,"tag":56,"props":65,"children":67},{"className":66},[],[68],{"type":49,"value":69},"commerce-common\u002Fcommerce_common\u002F",{"type":49,"value":71},", ",{"type":43,"tag":56,"props":73,"children":75},{"className":74},[],[76],{"type":49,"value":77},"shopping_agent\u002F",{"type":49,"value":63},{"type":43,"tag":56,"props":80,"children":82},{"className":81},[],[83],{"type":49,"value":84},"shopping-agent\u002Fcore\u002Fshopping_agent\u002F",{"type":49,"value":86},", and\n",{"type":43,"tag":56,"props":88,"children":90},{"className":89},[],[91],{"type":49,"value":92},"merchant_agent\u002F",{"type":49,"value":63},{"type":43,"tag":56,"props":95,"children":97},{"className":96},[],[98],{"type":49,"value":99},"merchant-agent\u002Fcore\u002Fmerchant_agent\u002F",{"type":49,"value":101},".",{"type":43,"tag":103,"props":104,"children":106},"h2",{"id":105},"the-three-breakpoints",[107],{"type":49,"value":108},"The three breakpoints",{"type":43,"tag":52,"props":110,"children":111},{},[112,114,120,122,128,130,136],{"type":49,"value":113},"A request's cacheable prefix runs ",{"type":43,"tag":56,"props":115,"children":117},{"className":116},[],[118],{"type":49,"value":119},"tools",{"type":49,"value":121},", then ",{"type":43,"tag":56,"props":123,"children":125},{"className":124},[],[126],{"type":49,"value":127},"system",{"type":49,"value":129},", then the messages; a breakpoint ends a span the next call reads\nback. The reference agents place three, all from ",{"type":43,"tag":56,"props":131,"children":133},{"className":132},[],[134],{"type":49,"value":135},"commerce_common\u002Fprompt_assembly.py",{"type":49,"value":137},":",{"type":43,"tag":139,"props":140,"children":141},"ol",{},[142,161,178],{"type":43,"tag":143,"props":144,"children":145},"li",{},[146,152,154,160],{"type":43,"tag":147,"props":148,"children":149},"strong",{},[150],{"type":49,"value":151},"The last tool",{"type":49,"value":153},": ",{"type":43,"tag":56,"props":155,"children":157},{"className":156},[],[158],{"type":49,"value":159},"with_tool_cache_control(tools)",{"type":49,"value":101},{"type":43,"tag":143,"props":162,"children":163},{},[164,169,170,176],{"type":43,"tag":147,"props":165,"children":166},{},[167],{"type":49,"value":168},"The static system text",{"type":49,"value":153},{"type":43,"tag":56,"props":171,"children":173},{"className":172},[],[174],{"type":49,"value":175},"build_system_blocks(static, context)",{"type":49,"value":177}," marks it and appends the context block unmarked.",{"type":43,"tag":143,"props":179,"children":180},{},[181,186,187,193],{"type":43,"tag":147,"props":182,"children":183},{},[184],{"type":49,"value":185},"The newest persisted message",{"type":49,"value":153},{"type":43,"tag":56,"props":188,"children":190},{"className":189},[],[191],{"type":49,"value":192},"build_request_messages(messages, rolling_breakpoint=...)",{"type":49,"value":194}," marks the last persisted\nblock on the outgoing request only; the next call then reads the earlier rounds, search payloads included, from cache.",{"type":43,"tag":52,"props":196,"children":197},{},[198,200,206,208,214],{"type":49,"value":199},"Everything per request goes in the context block, which the role's ",{"type":43,"tag":56,"props":201,"children":203},{"className":202},[],[204],{"type":49,"value":205},"build_dynamic_context(...)",{"type":49,"value":207}," renders once per turn.\nIn the static block or the tool list, those bytes would break the system or tool span on every request; in the context block they cost\none re-read of the conversation on a turn whose cart, page, or facts moved, and the rolling marker holds from that turn's\nnext round. The clock renders to the hour (",{"type":43,"tag":56,"props":209,"children":211},{"className":210},[],[212],{"type":49,"value":213},"context_clock",{"type":49,"value":215},"), so a new minute moves nothing.",{"type":43,"tag":103,"props":217,"children":219},{"id":218},"the-split-as-implemented",[220],{"type":49,"value":221},"The split as implemented",{"type":43,"tag":223,"props":224,"children":225},"ul",{},[226,245,255,321,363],{"type":43,"tag":143,"props":227,"children":228},{},[229,235,237,243],{"type":43,"tag":56,"props":230,"children":232},{"className":231},[],[233],{"type":49,"value":234},"build_static_system(config, skills)",{"type":49,"value":236}," in each role's ",{"type":43,"tag":56,"props":238,"children":240},{"className":239},[],[241],{"type":49,"value":242},"prompt.py",{"type":49,"value":244}," renders identity, the most-turns rules, the fence\nnotice, and the skill index from the config and the installed skills: the same bytes for the life of a deployment.",{"type":43,"tag":143,"props":246,"children":247},{},[248,253],{"type":43,"tag":56,"props":249,"children":251},{"className":250},[],[252],{"type":49,"value":205},{"type":49,"value":254}," in the same module renders the per-request material inside the role's data fence: shopping\ntakes preferences, memory facts, the cart, the page, the account block, and the time; merchant takes the store context,\nmemory facts, and the time. Backend context blocks have their own size cap.",{"type":43,"tag":143,"props":256,"children":257},{},[258,264,266,272,274,280,282,288,290,296,297,303,305,311,313,319],{"type":43,"tag":56,"props":259,"children":261},{"className":260},[],[262],{"type":49,"value":263},"ShoppingAgent.__init__",{"type":49,"value":265}," and ",{"type":43,"tag":56,"props":267,"children":269},{"className":268},[],[270],{"type":49,"value":271},"MerchantAgent.__init__",{"type":49,"value":273}," (each role's ",{"type":43,"tag":56,"props":275,"children":277},{"className":276},[],[278],{"type":49,"value":279},"runtime-messages-api",{"type":49,"value":281}," ",{"type":43,"tag":56,"props":283,"children":285},{"className":284},[],[286],{"type":49,"value":287},"orchestrator.py",{"type":49,"value":289},") build\n",{"type":43,"tag":56,"props":291,"children":293},{"className":292},[],[294],{"type":49,"value":295},"_static_system",{"type":49,"value":265},{"type":43,"tag":56,"props":298,"children":300},{"className":299},[],[301],{"type":49,"value":302},"_tools",{"type":49,"value":304}," once per process; ",{"type":43,"tag":56,"props":306,"children":308},{"className":307},[],[309],{"type":49,"value":310},"stream_turn",{"type":49,"value":312}," renders the context block per turn and calls\n",{"type":43,"tag":56,"props":314,"children":316},{"className":315},[],[317],{"type":49,"value":318},"build_request_messages",{"type":49,"value":320}," per model call.",{"type":43,"tag":143,"props":322,"children":323},{},[324,330,331,337,339,345,347,353,355,361],{"type":43,"tag":56,"props":325,"children":327},{"className":326},[],[328],{"type":49,"value":329},"build_tools",{"type":49,"value":236},{"type":43,"tag":56,"props":332,"children":334},{"className":333},[],[335],{"type":49,"value":336},"tools\u002Fregistry.py",{"type":49,"value":338}," emits the built-ins in a fixed order (the merchant builder adds ",{"type":43,"tag":56,"props":340,"children":342},{"className":341},[],[343],{"type":49,"value":344},"run_analysis",{"type":49,"value":346}," after them when enabled), the ",{"type":43,"tag":56,"props":348,"children":350},{"className":349},[],[351],{"type":49,"value":352},"load_skill",{"type":49,"value":354}," enum sorted,\nthen extensions in the order given, then ",{"type":43,"tag":56,"props":356,"children":358},{"className":357},[],[359],{"type":49,"value":360},"web_search",{"type":49,"value":362}," when enabled; every registered tool ships on every request, and\nthe executor decides on arrival whether a call can be served.",{"type":43,"tag":143,"props":364,"children":365},{},[366,368,374,376,382,384,390],{"type":49,"value":367},"The Agent SDK runtimes pass the same static text (plus ",{"type":43,"tag":56,"props":369,"children":371},{"className":370},[],[372],{"type":49,"value":373},"SKILL_TOOL_ADAPTER",{"type":49,"value":375},") and the same contracts, and the SDK caches\nthem; the hosted manifests carry the text as ",{"type":43,"tag":56,"props":377,"children":379},{"className":378},[],[380],{"type":49,"value":381},"system.md",{"type":49,"value":383},", which ",{"type":43,"tag":56,"props":385,"children":387},{"className":386},[],[388],{"type":49,"value":389},"scripts\u002Fcheck.py",{"type":49,"value":391}," compares with the builder.",{"type":43,"tag":103,"props":393,"children":395},{"id":394},"when-the-rolling-marker-is-skipped",[396],{"type":49,"value":397},"When the rolling marker is skipped",{"type":43,"tag":52,"props":399,"children":400},{},[401,403,409,411,417,419,424,426,431,433,439,441,447],{"type":49,"value":402},"Two rounds carry no marker. A bare first call (one message): a one-shot session would pay the write without a read, and\nthe second call's marker covers the first message anyway. A round whose ",{"type":43,"tag":56,"props":404,"children":406},{"className":405},[],[407],{"type":49,"value":408},"tool_choice",{"type":49,"value":410}," is other than ",{"type":43,"tag":56,"props":412,"children":414},{"className":413},[],[415],{"type":49,"value":416},"auto",{"type":49,"value":418}," (the\ngrounding-forced first iteration and the forced-text last one): ",{"type":43,"tag":56,"props":420,"children":422},{"className":421},[],[423],{"type":49,"value":408},{"type":49,"value":425}," keys the messages span, so an entry written\nunder a forced round is unreadable by the auto rounds after it; the system and tool spans still hit. Both live in\n",{"type":43,"tag":56,"props":427,"children":429},{"className":428},[],[430],{"type":49,"value":318},{"type":49,"value":432}," and the loops' ",{"type":43,"tag":56,"props":434,"children":436},{"className":435},[],[437],{"type":49,"value":438},"rolling_breakpoint=",{"type":49,"value":440}," argument; ",{"type":43,"tag":56,"props":442,"children":444},{"className":443},[],[445],{"type":49,"value":446},"rolling_conversation_cache",{"type":49,"value":448}," turns the marker off\nfor debugging.",{"type":43,"tag":103,"props":450,"children":452},{"id":451},"when-the-conversation-is-compacted",[453],{"type":49,"value":454},"When the conversation is compacted",{"type":43,"tag":52,"props":456,"children":457},{},[458,460,466,468,474,476,482,484,490],{"type":49,"value":459},"When a turn's last call was given ",{"type":43,"tag":56,"props":461,"children":463},{"className":462},[],[464],{"type":49,"value":465},"compact_history_above_tokens",{"type":49,"value":467}," or more (its usage says; the default is the platform's\nown tool-result-clearing default, a tenth of the window), the turn ends with ",{"type":43,"tag":56,"props":469,"children":471},{"className":470},[],[472],{"type":49,"value":473},"compact_history",{"type":49,"value":475}," in\n",{"type":43,"tag":56,"props":477,"children":479},{"className":478},[],[480],{"type":49,"value":481},"commerce_common\u002Fturn.py",{"type":49,"value":483},", which replaces the oldest tool results in the stored conversation with a one-line marker\nuntil the conversation is half its previous size. The next turn's first round rewrites the messages span once and later\nrounds read the shorter one; ",{"type":43,"tag":56,"props":485,"children":487},{"className":486},[],[488],{"type":49,"value":489},"turn_complete.results_cleared",{"type":49,"value":491}," tells a host that appends its transcript to rewrite it. The\nsystem and tool spans, the messages, and the write gates are unaffected; provenance lives on the session state.",{"type":43,"tag":103,"props":493,"children":495},{"id":494},"config-fields-that-are-prompt-bytes",[496],{"type":49,"value":497},"Config fields that are prompt bytes",{"type":43,"tag":52,"props":499,"children":500},{},[501,503,509,511,517,519,525,527,533],{"type":49,"value":502},"The fields marked ",{"type":43,"tag":56,"props":504,"children":506},{"className":505},[],[507],{"type":49,"value":508},"(prompt)",{"type":49,"value":510}," in ",{"type":43,"tag":56,"props":512,"children":514},{"className":513},[],[515],{"type":49,"value":516},"commerce_common\u002Fconfig.py",{"type":49,"value":518}," and the role configs, plus ",{"type":43,"tag":56,"props":520,"children":522},{"className":521},[],[523],{"type":49,"value":524},"enable_analysis",{"type":49,"value":526}," and\n",{"type":43,"tag":56,"props":528,"children":530},{"className":529},[],[531],{"type":49,"value":532},"max_items_per_change",{"type":49,"value":534},", which the merchant tool builder reads, render into the static text or the tool list; changing one is a redeploy and a miss on the next request:",{"type":43,"tag":536,"props":537,"children":538},"table",{},[539,558],{"type":43,"tag":540,"props":541,"children":542},"thead",{},[543],{"type":43,"tag":544,"props":545,"children":546},"tr",{},[547,553],{"type":43,"tag":548,"props":549,"children":550},"th",{},[551],{"type":49,"value":552},"Config",{"type":43,"tag":548,"props":554,"children":555},{},[556],{"type":49,"value":557},"Fields",{"type":43,"tag":559,"props":560,"children":561},"tbody",{},[562,614,679],{"type":43,"tag":544,"props":563,"children":564},{},[565,584],{"type":43,"tag":566,"props":567,"children":568},"td",{},[569,575,577,582],{"type":43,"tag":56,"props":570,"children":572},{"className":571},[],[573],{"type":49,"value":574},"BaseAgentConfig",{"type":49,"value":576}," (",{"type":43,"tag":56,"props":578,"children":580},{"className":579},[],[581],{"type":49,"value":516},{"type":49,"value":583},")",{"type":43,"tag":566,"props":585,"children":586},{},[587,593,594,600,601,607,608],{"type":43,"tag":56,"props":588,"children":590},{"className":589},[],[591],{"type":49,"value":592},"brand_name",{"type":49,"value":71},{"type":43,"tag":56,"props":595,"children":597},{"className":596},[],[598],{"type":49,"value":599},"assistant_name",{"type":49,"value":71},{"type":43,"tag":56,"props":602,"children":604},{"className":603},[],[605],{"type":49,"value":606},"brand_voice",{"type":49,"value":71},{"type":43,"tag":56,"props":609,"children":611},{"className":610},[],[612],{"type":49,"value":613},"enable_web_search",{"type":43,"tag":544,"props":615,"children":616},{},[617,634],{"type":43,"tag":566,"props":618,"children":619},{},[620,626,627,633],{"type":43,"tag":56,"props":621,"children":623},{"className":622},[],[624],{"type":49,"value":625},"ShoppingAgentConfig",{"type":49,"value":576},{"type":43,"tag":56,"props":628,"children":630},{"className":629},[],[631],{"type":49,"value":632},"shopping_agent\u002Fconfig.py",{"type":49,"value":583},{"type":43,"tag":566,"props":635,"children":636},{},[637,643,644,650,652,658,659,665,666,672,673],{"type":43,"tag":56,"props":638,"children":640},{"className":639},[],[641],{"type":49,"value":642},"domain_search_notes",{"type":49,"value":71},{"type":43,"tag":56,"props":645,"children":647},{"className":646},[],[648],{"type":49,"value":649},"enable_disclosures",{"type":49,"value":651},", and the system switches ",{"type":43,"tag":56,"props":653,"children":655},{"className":654},[],[656],{"type":49,"value":657},"enable_cart",{"type":49,"value":71},{"type":43,"tag":56,"props":660,"children":662},{"className":661},[],[663],{"type":49,"value":664},"enable_orders",{"type":49,"value":71},{"type":43,"tag":56,"props":667,"children":669},{"className":668},[],[670],{"type":49,"value":671},"enable_policies",{"type":49,"value":71},{"type":43,"tag":56,"props":674,"children":676},{"className":675},[],[677],{"type":49,"value":678},"enable_fulfillment",{"type":43,"tag":544,"props":680,"children":681},{},[682,699],{"type":43,"tag":566,"props":683,"children":684},{},[685,691,692,698],{"type":43,"tag":56,"props":686,"children":688},{"className":687},[],[689],{"type":49,"value":690},"MerchantAgentConfig",{"type":49,"value":576},{"type":43,"tag":56,"props":693,"children":695},{"className":694},[],[696],{"type":49,"value":697},"merchant_agent\u002Fconfig.py",{"type":49,"value":583},{"type":43,"tag":566,"props":700,"children":701},{},[702,707,708,714,715,721,722,728,729,735,736,742,743,749,750,756,758,763,764,770,772,778,779,785],{"type":43,"tag":56,"props":703,"children":705},{"className":704},[],[706],{"type":49,"value":524},{"type":49,"value":71},{"type":43,"tag":56,"props":709,"children":711},{"className":710},[],[712],{"type":49,"value":713},"require_host_approval",{"type":49,"value":71},{"type":43,"tag":56,"props":716,"children":718},{"className":717},[],[719],{"type":49,"value":720},"approval_surface",{"type":49,"value":71},{"type":43,"tag":56,"props":723,"children":725},{"className":724},[],[726],{"type":49,"value":727},"stage_shows_preview",{"type":49,"value":651},{"type":43,"tag":56,"props":730,"children":732},{"className":731},[],[733],{"type":49,"value":734},"enable_listing_edits",{"type":49,"value":71},{"type":43,"tag":56,"props":737,"children":739},{"className":738},[],[740],{"type":49,"value":741},"enable_inventory",{"type":49,"value":71},{"type":43,"tag":56,"props":744,"children":746},{"className":745},[],[747],{"type":49,"value":748},"enable_pricing",{"type":49,"value":71},{"type":43,"tag":56,"props":751,"children":753},{"className":752},[],[754],{"type":49,"value":755},"enable_campaigns",{"type":49,"value":757},"; ",{"type":43,"tag":56,"props":759,"children":761},{"className":760},[],[762],{"type":49,"value":532},{"type":49,"value":265},{"type":43,"tag":56,"props":765,"children":767},{"className":766},[],[768],{"type":49,"value":769},"max_search_results",{"type":49,"value":771}," set ",{"type":43,"tag":56,"props":773,"children":775},{"className":774},[],[776],{"type":49,"value":777},"maxItems",{"type":49,"value":265},{"type":43,"tag":56,"props":780,"children":782},{"className":781},[],[783],{"type":49,"value":784},"maximum",{"type":49,"value":786}," on tool schemas",{"type":43,"tag":52,"props":788,"children":789},{},[790,792,798,799,804,805,811,813,819,821,826,828,834],{"type":49,"value":791},"Skills, presentation extensions, and delegates are prompt bytes too. Gate lexicons, guardrail limits, memory settings,\nthe latency knobs (",{"type":43,"tag":56,"props":793,"children":795},{"className":794},[],[796],{"type":49,"value":797},"eager_tool_dispatch",{"type":49,"value":71},{"type":43,"tag":56,"props":800,"children":802},{"className":801},[],[803],{"type":49,"value":446},{"type":49,"value":71},{"type":43,"tag":56,"props":806,"children":808},{"className":807},[],[809],{"type":49,"value":810},"eager_partial_frames",{"type":49,"value":812},",\n",{"type":43,"tag":56,"props":814,"children":816},{"className":815},[],[817],{"type":49,"value":818},"close_on_presentation",{"type":49,"value":820},"), cart caps, and ",{"type":43,"tag":56,"props":822,"children":824},{"className":823},[],[825],{"type":49,"value":465},{"type":49,"value":827}," are not; ",{"type":43,"tag":56,"props":829,"children":831},{"className":830},[],[832],{"type":49,"value":833},"tests\u002Ftest_role_registries.py",{"type":49,"value":835},"\nasserts it for each.",{"type":43,"tag":103,"props":837,"children":839},{"id":838},"what-breaks-a-hit",[840],{"type":49,"value":841},"What breaks a hit",{"type":43,"tag":223,"props":843,"children":844},{},[845,850,855,866,871,876,888,893,898],{"type":43,"tag":143,"props":846,"children":847},{},[848],{"type":49,"value":849},"Anything per request in the static block or the tool list: a name, a cart count, a page, a clock, a request id. The\ncontext block takes the first four; a request id belongs nowhere.",{"type":43,"tag":143,"props":851,"children":852},{},[853],{"type":49,"value":854},"A set or dict iterated into prompt text or a schema without sorting.",{"type":43,"tag":143,"props":856,"children":857},{},[858,864],{"type":43,"tag":56,"props":859,"children":861},{"className":860},[],[862],{"type":49,"value":863},"tools[]",{"type":49,"value":865}," membership decided per request (a flag read per request instead of once at construction).",{"type":43,"tag":143,"props":867,"children":868},{},[869],{"type":49,"value":870},"Rebuilding the static text or the tool list inside the turn instead of in the constructor.",{"type":43,"tag":143,"props":872,"children":873},{},[874],{"type":49,"value":875},"Persisting the rolling marker into the stored conversation, which then gains one marker per turn.",{"type":43,"tag":143,"props":877,"children":878},{},[879,881,886],{"type":49,"value":880},"Toggling a ",{"type":43,"tag":56,"props":882,"children":884},{"className":883},[],[885],{"type":49,"value":508},{"type":49,"value":887}," field, a skill, or an extension on a running deployment; each is a redeploy.",{"type":43,"tag":143,"props":889,"children":890},{},[891],{"type":49,"value":892},"A prompt variant chosen per request instead of per deployment.",{"type":43,"tag":143,"props":894,"children":895},{},[896],{"type":49,"value":897},"A prefix under the model's minimum cacheable length, which writes nothing.",{"type":43,"tag":143,"props":899,"children":900},{},[901,903,908,910,916],{"type":49,"value":902},"A forced ",{"type":43,"tag":56,"props":904,"children":906},{"className":905},[],[907],{"type":49,"value":408},{"type":49,"value":909}," (grounding first, ",{"type":43,"tag":56,"props":911,"children":913},{"className":912},[],[914],{"type":49,"value":915},"none",{"type":49,"value":917}," last) misses the messages span only; the loops skip the marker there.",{"type":43,"tag":103,"props":919,"children":921},{"id":920},"how-to-verify",[922],{"type":49,"value":923},"How to verify",{"type":43,"tag":223,"props":925,"children":926},{},[927,1005,1031],{"type":43,"tag":143,"props":928,"children":929},{},[930,936,938,944,945,951,952,957,959,965,967,973,975,981,982,988,990,995,997,1003],{"type":43,"tag":56,"props":931,"children":933},{"className":932},[],[934],{"type":49,"value":935},"turn_complete",{"type":49,"value":937}," carries ",{"type":43,"tag":56,"props":939,"children":941},{"className":940},[],[942],{"type":49,"value":943},"usage",{"type":49,"value":576},{"type":43,"tag":56,"props":946,"children":948},{"className":947},[],[949],{"type":49,"value":950},"usage_totals",{"type":49,"value":510},{"type":43,"tag":56,"props":953,"children":955},{"className":954},[],[956],{"type":49,"value":481},{"type":49,"value":958},", the turn's calls summed) and\n",{"type":43,"tag":56,"props":960,"children":962},{"className":961},[],[963],{"type":49,"value":964},"elapsed_ms",{"type":49,"value":966},"; every call also logs one line with the same counters and its own time (",{"type":43,"tag":56,"props":968,"children":970},{"className":969},[],[971],{"type":49,"value":972},"log_model_call",{"type":49,"value":974},", on\n",{"type":43,"tag":56,"props":976,"children":978},{"className":977},[],[979],{"type":49,"value":980},"shopping_agent_runtime.orchestrator",{"type":49,"value":265},{"type":43,"tag":56,"props":983,"children":985},{"className":984},[],[986],{"type":49,"value":987},"merchant_agent_runtime.orchestrator",{"type":49,"value":989},"). Zero cache reads on the second turn\nof a conversation means the prefix changed. The counters worth charting per config version are cache reads as a\nshare of input, ",{"type":43,"tag":56,"props":991,"children":993},{"className":992},[],[994],{"type":49,"value":964},{"type":49,"value":996},", rounds per turn, and blocked ",{"type":43,"tag":56,"props":998,"children":1000},{"className":999},[],[1001],{"type":49,"value":1002},"tool_result",{"type":49,"value":1004}," events by gate.",{"type":43,"tag":143,"props":1006,"children":1007},{},[1008,1013,1015,1021,1023,1029],{"type":43,"tag":56,"props":1009,"children":1011},{"className":1010},[],[1012],{"type":49,"value":833},{"type":49,"value":1014}," builds each role's prompt and tools twice and compares bytes, checks the cache marks,\nand checks that non-prompt settings change nothing; ",{"type":43,"tag":56,"props":1016,"children":1018},{"className":1017},[],[1019],{"type":49,"value":1020},"commerce-common\u002Ftests\u002Ftest_prompt_assembly.py",{"type":49,"value":1022}," covers the block\nbuilders, the context clock, the rolling marker, and the skip cases; ",{"type":43,"tag":56,"props":1024,"children":1026},{"className":1025},[],[1027],{"type":49,"value":1028},"tests\u002Ftest_turn_loop.py",{"type":49,"value":1030}," pins both loops to the\nsame blocks, marker, and clean history. A deployment's tests carry the same three checks.",{"type":43,"tag":143,"props":1032,"children":1033},{},[1034],{"type":49,"value":1035},"In a deployed environment, run one three-turn conversation and read the second and third turns' usage; a proxy or retry\nlayer that rewrites requests shows up here and nowhere else.",{"items":1037,"total":1141},[1038,1055,1070,1085,1100,1107,1124],{"slug":1039,"name":1039,"fn":1040,"description":1041,"org":1042,"tags":1043,"stars":26,"repoUrl":27,"updatedAt":1054},"catalog-listings","create and improve e-commerce listing content","Creating and improving listing content, covering titles, descriptions, attribute completeness, categorization fixes, image callouts written as text, edits written from material the operator supplies, and content-quality audits and bulk fixes across many listings. Not needed for questions about how a listing is performing.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1044,1047,1050,1051],{"name":1045,"slug":1046,"type":16},"Content Creation","content-creation",{"name":1048,"slug":1049,"type":16},"Content Strategy","content-strategy",{"name":21,"slug":22,"type":16},{"name":1052,"slug":1053,"type":16},"Marketing","marketing","2026-09-04T08:00:35.843558",{"slug":1056,"name":1056,"fn":1057,"description":1058,"org":1059,"tags":1060,"stars":26,"repoUrl":27,"updatedAt":1069},"commerce-architecture","design and review commerce agent architecture","How the reference commerce agents of either role are structured, covering the loop, where each rule lives, skills, the backend interface, delegates, and model fields. Load when designing or reviewing the structure of a shopping or merchant agent.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1061,1062,1065,1066],{"name":24,"slug":25,"type":16},{"name":1063,"slug":1064,"type":16},"Architecture","architecture",{"name":21,"slug":22,"type":16},{"name":1067,"slug":1068,"type":16},"Engineering","engineering","2026-09-04T08:00:39.152322",{"slug":1071,"name":1071,"fn":1072,"description":1073,"org":1074,"tags":1075,"stars":26,"repoUrl":27,"updatedAt":1084},"commerce-evals","run behavioral evals for commerce agents","Authoring and running behavioral evals for a shopping or merchant agent, covering the case shape, authoring rules, code graders and judges, the run pattern, and poisoned fixtures. Load when writing eval cases or rubrics or deciding how a suite runs.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1076,1077,1078,1081],{"name":24,"slug":25,"type":16},{"name":21,"slug":22,"type":16},{"name":1079,"slug":1080,"type":16},"Evals","evals",{"name":1082,"slug":1083,"type":16},"Testing","testing","2026-09-04T08:00:36.204114",{"slug":1086,"name":1086,"fn":1087,"description":1088,"org":1089,"tags":1090,"stars":26,"repoUrl":27,"updatedAt":1099},"commerce-merchant-operations","manage merchant operations and store rules","The reference merchant agent, covering its flows, staged changes and host approval, metrics grounding, the analysis delegate, store memory, components, and marketplace rules. Load when building or reviewing an agent for the operators of a store, property, subscriber base, or venue.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1091,1092,1093,1096],{"name":24,"slug":25,"type":16},{"name":21,"slug":22,"type":16},{"name":1094,"slug":1095,"type":16},"Management","management",{"name":1097,"slug":1098,"type":16},"Operations","operations","2026-09-04T08:00:35.464555",{"slug":4,"name":4,"fn":5,"description":6,"org":1101,"tags":1102,"stars":26,"repoUrl":27,"updatedAt":28},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1103,1104,1105,1106],{"name":24,"slug":25,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":18,"slug":19,"type":16},{"slug":1108,"name":1108,"fn":1109,"description":1110,"org":1111,"tags":1112,"stars":26,"repoUrl":27,"updatedAt":1123},"commerce-trust-safety","enforce trust and safety rules","The rules the reference agents enforce in code for third-party content, writes, grounding, identity, and memory, each with its module, plus two adversarial-eval rules. Load when handling untrusted tool results, guarding a write tool, or scoping what an agent remembers.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1113,1114,1117,1120],{"name":24,"slug":25,"type":16},{"name":1115,"slug":1116,"type":16},"Compliance","compliance",{"name":1118,"slug":1119,"type":16},"Memory","memory",{"name":1121,"slug":1122,"type":16},"Security","security","2026-09-04T08:00:42.609563",{"slug":1125,"name":1125,"fn":1126,"description":1127,"org":1128,"tags":1129,"stars":26,"repoUrl":27,"updatedAt":1140},"commerce-ui-tools","implement commerce agent UI components","The reference presentation-tool contract, covering server-side enrichment, suggestion chips, the event stream, progressive rendering, both roles' built-in components, and adding a vertical component. Load when building or reviewing how a commerce agent's output reaches a frontend.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1130,1133,1134,1137],{"name":1131,"slug":1132,"type":16},"Design","design",{"name":21,"slug":22,"type":16},{"name":1135,"slug":1136,"type":16},"Frontend","frontend",{"name":1138,"slug":1139,"type":16},"UI Components","ui-components","2026-09-04T08:00:38.801247",16,{"items":1143,"total":1321},[1144,1160,1177,1191,1203,1220,1236,1247,1268,1284,1298,1313],{"slug":1145,"name":1145,"fn":1146,"description":1147,"org":1148,"tags":1149,"stars":1157,"repoUrl":1158,"updatedAt":1159},"academy-guide","recommend Claude Academy resources","Stop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: \"how do I\", \"how can I\", \"getting started with\", \"what can Claude do\", \"teach me\", \"learn to use\"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1150,1151,1154],{"name":9,"slug":8,"type":16},{"name":1152,"slug":1153,"type":16},"Documentation","documentation",{"name":1155,"slug":1156,"type":16},"Education","education",161831,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills","2026-08-19T03:59:01.021254",{"slug":1161,"name":1161,"fn":1162,"description":1163,"org":1164,"tags":1165,"stars":1157,"repoUrl":1158,"updatedAt":1176},"algorithmic-art","create algorithmic art with p5.js","Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1166,1169,1170,1173],{"name":1167,"slug":1168,"type":16},"Creative","creative",{"name":1131,"slug":1132,"type":16},{"name":1171,"slug":1172,"type":16},"Generative Art","generative-art",{"name":1174,"slug":1175,"type":16},"JavaScript","javascript","2026-04-06T17:56:15.455818",{"slug":1178,"name":1178,"fn":1179,"description":1180,"org":1181,"tags":1182,"stars":1157,"repoUrl":1158,"updatedAt":1190},"brand-guidelines","apply Anthropic brand colors and typography","Applies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1183,1186,1187],{"name":1184,"slug":1185,"type":16},"Branding","branding",{"name":1131,"slug":1132,"type":16},{"name":1188,"slug":1189,"type":16},"Typography","typography","2026-04-06T17:56:05.042852",{"slug":1192,"name":1192,"fn":1193,"description":1194,"org":1195,"tags":1196,"stars":1157,"repoUrl":1158,"updatedAt":1202},"canvas-design","create posters and visual art as PNG or PDF","Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1197,1198,1199],{"name":1167,"slug":1168,"type":16},{"name":1131,"slug":1132,"type":16},{"name":1200,"slug":1201,"type":16},"PDF","pdf","2026-04-06T17:56:03.794732",{"slug":1204,"name":1204,"fn":1205,"description":1206,"org":1207,"tags":1208,"stars":1157,"repoUrl":1158,"updatedAt":1219},"claude-api","build apps with the Claude API","Reference for the Claude API \u002F Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration.\nTRIGGER — read BEFORE opening the target file; don't skip because it \"looks like a one-liner\" — whenever: the prompt names Claude\u002FAnthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing\u002Fmodel choice\u002Flimits\u002Fcaching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent\u002FMCP\u002Ftool-definition\u002Fmulti-agent\u002FRAG\u002FLLM-judge\u002Fcomputer-use; generate\u002Fsummarize\u002Fextract\u002Fclassify\u002Frewrite\u002Fconverse over NL; debugging refusals\u002Fcutoffs\u002Fstreaming\u002Ftool-calls\u002Ftokens).\nSKIP only when another provider is being worked on (overrides all triggers): OpenAI\u002FGPT\u002FGemini\u002FLlama\u002FMistral\u002FCohere\u002FOllama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1209,1210,1211,1214,1216],{"name":24,"slug":25,"type":16},{"name":9,"slug":8,"type":16},{"name":1212,"slug":1213,"type":16},"Anthropic SDK","anthropic-sdk",{"name":1215,"slug":1204,"type":16},"Claude API",{"name":1217,"slug":1218,"type":16},"LLM","llm","2026-09-04T07:28:24.898544",{"slug":1221,"name":1221,"fn":1222,"description":1223,"org":1224,"tags":1225,"stars":1157,"repoUrl":1158,"updatedAt":1235},"discernment-nudge","provide discernment nudges for user decisions","After you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1226,1229,1232],{"name":1227,"slug":1228,"type":16},"Coaching","coaching",{"name":1230,"slug":1231,"type":16},"Productivity","productivity",{"name":1233,"slug":1234,"type":16},"Strategy","strategy","2026-08-19T03:59:01.539281",{"slug":1237,"name":1237,"fn":1238,"description":1239,"org":1240,"tags":1241,"stars":1157,"repoUrl":1158,"updatedAt":1246},"doc-coauthoring","co-author documentation and technical specs","Guide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1242,1243],{"name":1152,"slug":1153,"type":16},{"name":1244,"slug":1245,"type":16},"Technical Writing","technical-writing","2026-04-06T17:56:14.18897",{"slug":1248,"name":1248,"fn":1249,"description":1250,"org":1251,"tags":1252,"stars":1157,"repoUrl":1158,"updatedAt":1267},"docx","create and edit Word documents","Use this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1253,1256,1258,1261,1264],{"name":1254,"slug":1255,"type":16},"Documents","documents",{"name":1257,"slug":1248,"type":16},"DOCX",{"name":1259,"slug":1260,"type":16},"Office","office",{"name":1262,"slug":1263,"type":16},"Templates","templates",{"name":1265,"slug":1266,"type":16},"Word","word","2026-07-18T05:16:23.136271",{"slug":1269,"name":1269,"fn":1270,"description":1271,"org":1272,"tags":1273,"stars":1157,"repoUrl":1158,"updatedAt":1283},"frontend-design","design production-grade frontend interfaces","Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1274,1275,1276,1279,1282],{"name":1131,"slug":1132,"type":16},{"name":1135,"slug":1136,"type":16},{"name":1277,"slug":1278,"type":16},"React","react",{"name":1280,"slug":1281,"type":16},"Tailwind CSS","tailwind-css",{"name":1138,"slug":1139,"type":16},"2026-09-04T07:28:23.795756",{"slug":1285,"name":1285,"fn":1286,"description":1287,"org":1288,"tags":1289,"stars":1157,"repoUrl":1158,"updatedAt":1297},"internal-comms","write internal company communications","A set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Claude should use this skill whenever asked to write some sort of internal communications (status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc.).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1290,1293,1294],{"name":1291,"slug":1292,"type":16},"Communications","communications",{"name":1262,"slug":1263,"type":16},{"name":1295,"slug":1296,"type":16},"Writing","writing","2026-04-06T17:56:20.695522",{"slug":1299,"name":1299,"fn":1300,"description":1301,"org":1302,"tags":1303,"stars":1157,"repoUrl":1158,"updatedAt":1312},"mcp-builder","build MCP servers","Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node\u002FTypeScript (MCP SDK).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1304,1305,1308,1309],{"name":24,"slug":25,"type":16},{"name":1306,"slug":1307,"type":16},"API Development","api-development",{"name":1217,"slug":1218,"type":16},{"name":1310,"slug":1311,"type":16},"MCP","mcp","2026-04-06T17:56:10.357665",{"slug":1201,"name":1201,"fn":1314,"description":1315,"org":1316,"tags":1317,"stars":1157,"repoUrl":1158,"updatedAt":1320},"read edit and manipulate PDF files","Use this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text\u002Ftables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting\u002Fdecrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1318,1319],{"name":1254,"slug":1255,"type":16},{"name":1200,"slug":1201,"type":16},"2026-04-06T17:56:02.483316",521]