[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DBR3SCVKDWEX96Q95SCX56":3,"skill-same:01M7DBR3SCVKDWEX96Q95SCX56":47},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":14,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M7DBR3SCVKDWEX96Q95SCX56","authoring-pi-extensions",null,"Write or change a Pi extension against the current public API — the extension entry shape, how package.json discovery fields are glob-expanded, the full lifecycle event list and what each handler may return, how before_agent_start chains the system prompt, how resources_discover contributes skills and prompts, and which surfaces are current rather than deprecated. Load before adding an event handler, command, flag, tool, or renderer to a Pi extension, and before trusting recollection of an event name or return shape.","CenterForAgenticAI\u002Ftools","CenterForAgenticAI","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F250424195?v=4","01M7DBNBBRNZ6NT6WYSEPYK811","CAAIR","01M7DBNBCBH3B7HHWP0ACQW3F7",0,"MIT","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-dev\u002Fskills\u002Fauthoring-pi-extensions\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: authoring-pi-extensions\ndescription: Write or change a Pi extension against the current public API — the extension entry shape, how package.json discovery fields are glob-expanded, the full lifecycle event list and what each handler may return, how before_agent_start chains the system prompt, how resources_discover contributes skills and prompts, and which surfaces are current rather than deprecated. Load before adding an event handler, command, flag, tool, or renderer to a Pi extension, and before trusting recollection of an event name or return shape.\nload: recognition\n---\n\n# Authoring Pi extensions\n\nEverything here is read from the installed typings of\n`@earendil-works\u002Fpi-coding-agent`, at\n`node_modules\u002F@earendil-works\u002Fpi-coding-agent\u002Fdist\u002Fcore\u002Fextensions\u002Ftypes.d.ts`.\nWhen this skill and that file disagree, the file wins — say so rather than\nfollowing a stale instruction.\n\n## The entry shape\n\nAn extension is a module whose default export takes the extension API object.\n\n```ts\nimport type { ExtensionAPI } from \"@earendil-works\u002Fpi-coding-agent\";\n\nexport default function (pi: ExtensionAPI): void {\n  pi.registerCommand(\"example\", { \u002F* ... *\u002F });\n}\n```\n\nA factory that returns that function works equally well, and is worth preferring\nwhen the extension has injectable dependencies, because it gives tests a seam\nwithout a module mock.\n\n## Discovery is declared in package.json\n\nPi finds an extension through the `pi` field. All three paths are **glob\nexpanded**, so a directory pattern is legitimate:\n\n```json\n{\n  \"pi\": {\n    \"extensions\": [\".\u002Findex.ts\", \".\u002Fmodules\u002F*\u002Findex.ts\"],\n    \"skills\": [\".\u002Fskills\", \".\u002Fmodules\u002F*\u002Fskills\"],\n    \"prompts\": [\".\u002Fprompts\"]\n  }\n}\n```\n\n`skills` entries may be a `.md` file or a directory. Directories are recursively\nscanned for `SKILL.md`; a directory containing `SKILL.md` is a skill root, so\nthat subtree is not scanned further. `prompts` entries may be a `.md` file or a\ndirectory. Directories declared in `package.json` are recursively scanned for\n`.md` prompt templates. Paths returned from `resources_discover` are passed\ndirectly to the loaders: skill directories are recursive, while prompt\ndirectories are scanned **non-recursively** for `.md` files.\n\nThe trap worth knowing: **`pi.skills` is resolved unconditionally.** There is no\nper-flag or per-condition gate on that path. If a skill must only exist under\nsome condition, declaring it here is the wrong mechanism — contribute it from a\n`resources_discover` handler instead.\n\n## Registration surface\n\n| Call | Purpose |\n| --- | --- |\n| `registerCommand(name, options)` | A slash command. `options` is `RegisteredCommand` without `name` and `sourceInfo`. |\n| `registerFlag(name, options)` | A CLI flag. `options` is `{ description?, type: \"boolean\" \\| \"string\", default? }`. |\n| `getFlag(name)` | Reads it back: `boolean \\| string \\| undefined`. |\n| `registerTool(tool)` | A tool the model may call. |\n| `registerShortcut(key, options)` | A keyboard shortcut. |\n| `registerMessageRenderer(customType, renderer)` | Renders a custom message. |\n| `registerMarkdownTransformer(transformer)` | Synchronously transforms user, assistant, and thinking Markdown for display only. |\n| `registerEntryRenderer(customType, renderer)` | Renders a custom entry. Custom entries never enter model context. |\n| `sendMessage(message, options?)` | Sends a custom message into the session. |\n\nFlags share **one global namespace across every loaded extension**. A collision\nis diagnosed and precedence follows load order, so namespace anything that is\nnot obviously yours.\n\n## Lifecycle events\n\nRegister with `pi.on(event, handler)`. Only the events listed here exist; do not\ninvent neighbours by analogy.\n\nEvents that accept a **result** that changes behaviour:\n\n| Event | Result | Effect |\n| --- | --- | --- |\n| `project_trust` | `ProjectTrustEventResult` | Must return `trusted: \"yes\" | \"no\" | \"undecided\"`; the optional `remember?` field is available for the decision. |\n| `before_agent_start` | `BeforeAgentStartEventResult` | `systemPrompt` replaces the prompt for the turn; chains across extensions. Also accepts `message`. |\n| `input` | `InputEventResult` | `{ action: \"continue\" }`, `{ action: \"transform\", text, images? }`, or `{ action: \"handled\" }`. |\n| `context` | `ContextEventResult` | `messages` replaces the context messages. |\n| `tool_call` | `ToolCallEventResult` | `block?` can stop execution, `reason?` explains it, and `terminate?` can end the agent early only when every finalized result in the batch is terminating. To change arguments, mutate `event.input` in place instead. |\n| `tool_result` | `ToolResultEventResult` | `content?`, `details?`, `isError?`, and `usage?` can replace result fields. |\n| `message_end` | `MessageEndEventResult` | `message?` replaces the finalized message, and **must keep the original role**. |\n| `user_bash` | `UserBashEventResult` | `operations?` supplies execution, or `result?` replaces it entirely. |\n| `resources_discover` | `ResourcesDiscoverResult` | `skillPaths?`, `promptPaths?`, and `themePaths?` add resource paths. |\n| `session_before_compact` | `SessionBeforeCompactResult` | `cancel?` cancels; `compaction?` supplies a replacement compaction. |\n| `session_before_fork` | `SessionBeforeForkResult` | `cancel?` cancels; `skipConversationRestore?` controls conversation restoration. |\n| `session_before_switch` | `SessionBeforeSwitchResult` | `cancel?` cancels the switch. |\n| `session_before_tree` | `SessionBeforeTreeResult` | `cancel?`, `summary?: { summary: string, details?: unknown, usage?: Usage }`, `customInstructions?`, `replaceInstructions?`, and `label?` control navigation. |\n| `before_provider_request` | `BeforeProviderRequestEventResult` | An `unknown` return value can replace the request payload. |\n\nObservation-only events, where a return value changes nothing:\n\n`agent_start`, `agent_end`, `agent_settled`, `ui_prompt_start`,\n`ui_prompt_end`, `turn_start`, `turn_end`, `message_start`, `message_update`,\n`session_start`, `session_shutdown`,\n`session_compact`, `session_compact_failed`, `session_tree`, `session_info_changed`, `model_select`,\n`thinking_level_select`, `tool_execution_start`, `tool_execution_update`,\n`tool_execution_end`, `after_provider_response`, `before_provider_headers`.\n\n`ui_prompt_start` and `ui_prompt_end` bracket the outermost blocking extension\nUI prompt. Both report `reason: \"ui_prompt\"`, the prompt `kind`, and an optional\nhuman-facing `title`; nested prompts stay within the same outer pair.\n\n## Two contracts that are easy to get wrong\n\n**`before_agent_start` replaces, so you must compose.**\n`BeforeAgentStartEventResult.systemPrompt` is documented as replacing the prompt\nfor the turn, chaining when several extensions return one. Returning your own\ntext bare therefore discards whatever ran before you. Append to what you were\nhanded:\n\n```ts\npi.on(\"before_agent_start\", (event) => ({\n  systemPrompt: `${event.systemPrompt}\\n\\n${mySection}`,\n}));\n```\n\n**`resources_discover` fires at startup or reload.**\n\n```ts\nexport interface ResourcesDiscoverEvent {\n  type: \"resources_discover\";\n  cwd: string;\n  reason: \"startup\" | \"reload\";\n}\n```\n\nSkills and prompt templates cannot appear part-way through a session without a\nreload. Anything that can flip mid-session — a command toggle, say — must call\n`await ctx.reload()` before its contributed resources appear, while a\nstartup-time CLI flag can activate them during the initial load. Decide the\nactivation mechanism with that constraint in mind rather than discovering it\nafterwards.\n\n`session_start` carries a wider `reason`: `\"startup\" | \"reload\" | \"new\" |\n\"resume\" | \"fork\"`.\n\n## Current versus deprecated\n\nUse the public `@earendil-works\u002Fpi-*` packages. Do not import Pi implementation\nmodules by deep path, and do not use the deprecated `@mariozechner\u002F*` packages.\n\nTreat an optional integration as optional: load it dynamically, and degrade with\na clear message when it is absent, rather than making it a hard dependency.\n\n## Keep requirements separate from design\n\nDo not promote an implementation choice into a requirement. Before planning a\nchange, separate three sources of authority:\n\n- **source requirements** — behavior explicitly required by the request,\n  specification, or acceptance criteria;\n- **repository invariants** — rules the repository already establishes and must\n  keep true;\n- **proposed design choices** — mechanisms introduced by the implementer. They\n  remain optional unless an authoritative source accepts them.\n\nIf removing a proposed mechanism still meets every source requirement and\nrepository invariant, treat it as optional and prefer not to add it.\n\n### Stop before expanding the contract\n\nBefore adding cryptography, credentials, authorization, a new protocol, durable\nrecovery behavior, or a new trust boundary that no source requirement or\nrepository invariant requires:\n\n1. Show why the existing trusted path cannot satisfy acceptance.\n2. Name the concrete attacker and capability the stronger mechanism addresses.\n3. Compare it with the simplest sufficient design.\n4. Stop for human approval when the stronger contract would materially expand\n   the requested change.\n\nThis gate does not weaken a security mechanism required by a human decision,\nsource specification, or documented repository invariant. It interrupts only an\nunsupported expansion.\n\n### Review the requirement before the mechanism\n\nWhen reviewing a change, first ask whether each mechanism is required. Flag an\nunsupported invariant or disproportionate scope before treating an agent-proposed\nchoice as settled architecture. Report these separately:\n\n- a **source-contract failure**, where required behavior or security is wrong;\n- an **optional-defense failure**, where an unrequired added defense is broken;\n- **practical severity**, rated from realistic prerequisites and consequences.\n\nDo not turn “the optional defense is broken” into “the required behavior is\ninsecure.”\n\n## Parallel test execution and isolation\n\nWhole-repository gates run test files in parallel by default. Node's test\nrunner already supplies process-level concurrency when no serializing flag is\npresent, so do not add a serializing flag to the whole gate.\n\nIf a genuinely serial suite needs serialization, give it its own named serial\nscript and put a nearby written reason next to that script. Never serialize the whole\ngate for one suite. Keep coverage wrappers such as `c8` transparent to the\nunderlying runner's concurrency: they must preserve the concurrency selected by\nthe wrapped runner rather than adding their own throttle.\n\nCI may set an optional `TEST_CONCURRENCY` control to reduce CI pressure. Accept\nthe runner default when it is unset. Reject a fixed low ceiling that ignores the\ncapacity of the host running the tests.\n\nThe slowest file sets a lower bound on parallel wall time. Treat files\naround 30 seconds or slower as split candidates. Prefer fakes or event hooks\nover `setTimeout` or sleep waits so tests do not spend their time waiting on a\nclock.\n\nMake tests parallel-safe with unique temporary directories. Avoid fixed shared\npaths, socket names, ports, daemons, and other uncontrolled shared state.\nSerialization pressure should trigger an isolation review. Fix the shared\nresource rather than hiding the collision behind a serial gate.\n\n## Before you claim it works\n\nType-check against the real typings rather than trusting an event name that\nlooks plausible. An unknown event name is a silent no-op — the handler simply\nnever runs, and nothing reports it.\n","# Authoring Pi extensions\n\nEverything here is read from the installed typings of\n`@earendil-works\u002Fpi-coding-agent`, at\n`node_modules\u002F@earendil-works\u002Fpi-coding-agent\u002Fdist\u002Fcore\u002Fextensions\u002Ftypes.d.ts`.\nWhen this skill and that file disagree, the file wins — say so rather than\nfollowing a stale instruction.\n\n## The entry shape\n\nAn extension is a module whose default export takes the extension API object.\n\n```ts\nimport type { ExtensionAPI } from \"@earendil-works\u002Fpi-coding-agent\";\n\nexport default function (pi: ExtensionAPI): void {\n  pi.registerCommand(\"example\", { \u002F* ... *\u002F });\n}\n```\n\nA factory that returns that function works equally well, and is worth preferring\nwhen the extension has injectable dependencies, because it gives tests a seam\nwithout a module mock.\n\n## Discovery is declared in package.json\n\nPi finds an extension through the `pi` field. All three paths are **glob\nexpanded**, so a directory pattern is legitimate:\n\n```json\n{\n  \"pi\": {\n    \"extensions\": [\".\u002Findex.ts\", \".\u002Fmodules\u002F*\u002Findex.ts\"],\n    \"skills\": [\".\u002Fskills\", \".\u002Fmodules\u002F*\u002Fskills\"],\n    \"prompts\": [\".\u002Fprompts\"]\n  }\n}\n```\n\n`skills` entries may be a `.md` file or a directory. Directories are recursively\nscanned for `SKILL.md`; a directory containing `SKILL.md` is a skill root, so\nthat subtree is not scanned further. `prompts` entries may be a `.md` file or a\ndirectory. Directories declared in `package.json` are recursively scanned for\n`.md` prompt templates. Paths returned from `resources_discover` are passed\ndirectly to the loaders: skill directories are recursive, while prompt\ndirectories are scanned **non-recursively** for `.md` files.\n\nThe trap worth knowing: **`pi.skills` is resolved unconditionally.** There is no\nper-flag or per-condition gate on that path. If a skill must only exist under\nsome condition, declaring it here is the wrong mechanism — contribute it from a\n`resources_discover` handler instead.\n\n## Registration surface\n\n| Call | Purpose |\n| --- | --- |\n| `registerCommand(name, options)` | A slash command. `options` is `RegisteredCommand` without `name` and `sourceInfo`. |\n| `registerFlag(name, options)` | A CLI flag. `options` is `{ description?, type: \"boolean\" \\| \"string\", default? }`. |\n| `getFlag(name)` | Reads it back: `boolean \\| string \\| undefined`. |\n| `registerTool(tool)` | A tool the model may call. |\n| `registerShortcut(key, options)` | A keyboard shortcut. |\n| `registerMessageRenderer(customType, renderer)` | Renders a custom message. |\n| `registerMarkdownTransformer(transformer)` | Synchronously transforms user, assistant, and thinking Markdown for display only. |\n| `registerEntryRenderer(customType, renderer)` | Renders a custom entry. Custom entries never enter model context. |\n| `sendMessage(message, options?)` | Sends a custom message into the session. |\n\nFlags share **one global namespace across every loaded extension**. A collision\nis diagnosed and precedence follows load order, so namespace anything that is\nnot obviously yours.\n\n## Lifecycle events\n\nRegister with `pi.on(event, handler)`. Only the events listed here exist; do not\ninvent neighbours by analogy.\n\nEvents that accept a **result** that changes behaviour:\n\n| Event | Result | Effect |\n| --- | --- | --- |\n| `project_trust` | `ProjectTrustEventResult` | Must return `trusted: \"yes\" | \"no\" | \"undecided\"`; the optional `remember?` field is available for the decision. |\n| `before_agent_start` | `BeforeAgentStartEventResult` | `systemPrompt` replaces the prompt for the turn; chains across extensions. Also accepts `message`. |\n| `input` | `InputEventResult` | `{ action: \"continue\" }`, `{ action: \"transform\", text, images? }`, or `{ action: \"handled\" }`. |\n| `context` | `ContextEventResult` | `messages` replaces the context messages. |\n| `tool_call` | `ToolCallEventResult` | `block?` can stop execution, `reason?` explains it, and `terminate?` can end the agent early only when every finalized result in the batch is terminating. To change arguments, mutate `event.input` in place instead. |\n| `tool_result` | `ToolResultEventResult` | `content?`, `details?`, `isError?`, and `usage?` can replace result fields. |\n| `message_end` | `MessageEndEventResult` | `message?` replaces the finalized message, and **must keep the original role**. |\n| `user_bash` | `UserBashEventResult` | `operations?` supplies execution, or `result?` replaces it entirely. |\n| `resources_discover` | `ResourcesDiscoverResult` | `skillPaths?`, `promptPaths?`, and `themePaths?` add resource paths. |\n| `session_before_compact` | `SessionBeforeCompactResult` | `cancel?` cancels; `compaction?` supplies a replacement compaction. |\n| `session_before_fork` | `SessionBeforeForkResult` | `cancel?` cancels; `skipConversationRestore?` controls conversation restoration. |\n| `session_before_switch` | `SessionBeforeSwitchResult` | `cancel?` cancels the switch. |\n| `session_before_tree` | `SessionBeforeTreeResult` | `cancel?`, `summary?: { summary: string, details?: unknown, usage?: Usage }`, `customInstructions?`, `replaceInstructions?`, and `label?` control navigation. |\n| `before_provider_request` | `BeforeProviderRequestEventResult` | An `unknown` return value can replace the request payload. |\n\nObservation-only events, where a return value changes nothing:\n\n`agent_start`, `agent_end`, `agent_settled`, `ui_prompt_start`,\n`ui_prompt_end`, `turn_start`, `turn_end`, `message_start`, `message_update`,\n`session_start`, `session_shutdown`,\n`session_compact`, `session_compact_failed`, `session_tree`, `session_info_changed`, `model_select`,\n`thinking_level_select`, `tool_execution_start`, `tool_execution_update`,\n`tool_execution_end`, `after_provider_response`, `before_provider_headers`.\n\n`ui_prompt_start` and `ui_prompt_end` bracket the outermost blocking extension\nUI prompt. Both report `reason: \"ui_prompt\"`, the prompt `kind`, and an optional\nhuman-facing `title`; nested prompts stay within the same outer pair.\n\n## Two contracts that are easy to get wrong\n\n**`before_agent_start` replaces, so you must compose.**\n`BeforeAgentStartEventResult.systemPrompt` is documented as replacing the prompt\nfor the turn, chaining when several extensions return one. Returning your own\ntext bare therefore discards whatever ran before you. Append to what you were\nhanded:\n\n```ts\npi.on(\"before_agent_start\", (event) => ({\n  systemPrompt: `${event.systemPrompt}\\n\\n${mySection}`,\n}));\n```\n\n**`resources_discover` fires at startup or reload.**\n\n```ts\nexport interface ResourcesDiscoverEvent {\n  type: \"resources_discover\";\n  cwd: string;\n  reason: \"startup\" | \"reload\";\n}\n```\n\nSkills and prompt templates cannot appear part-way through a session without a\nreload. Anything that can flip mid-session — a command toggle, say — must call\n`await ctx.reload()` before its contributed resources appear, while a\nstartup-time CLI flag can activate them during the initial load. Decide the\nactivation mechanism with that constraint in mind rather than discovering it\nafterwards.\n\n`session_start` carries a wider `reason`: `\"startup\" | \"reload\" | \"new\" |\n\"resume\" | \"fork\"`.\n\n## Current versus deprecated\n\nUse the public `@earendil-works\u002Fpi-*` packages. Do not import Pi implementation\nmodules by deep path, and do not use the deprecated `@mariozechner\u002F*` packages.\n\nTreat an optional integration as optional: load it dynamically, and degrade with\na clear message when it is absent, rather than making it a hard dependency.\n\n## Keep requirements separate from design\n\nDo not promote an implementation choice into a requirement. Before planning a\nchange, separate three sources of authority:\n\n- **source requirements** — behavior explicitly required by the request,\n  specification, or acceptance criteria;\n- **repository invariants** — rules the repository already establishes and must\n  keep true;\n- **proposed design choices** — mechanisms introduced by the implementer. They\n  remain optional unless an authoritative source accepts them.\n\nIf removing a proposed mechanism still meets every source requirement and\nrepository invariant, treat it as optional and prefer not to add it.\n\n### Stop before expanding the contract\n\nBefore adding cryptography, credentials, authorization, a new protocol, durable\nrecovery behavior, or a new trust boundary that no source requirement or\nrepository invariant requires:\n\n1. Show why the existing trusted path cannot satisfy acceptance.\n2. Name the concrete attacker and capability the stronger mechanism addresses.\n3. Compare it with the simplest sufficient design.\n4. Stop for human approval when the stronger contract would materially expand\n   the requested change.\n\nThis gate does not weaken a security mechanism required by a human decision,\nsource specification, or documented repository invariant. It interrupts only an\nunsupported expansion.\n\n### Review the requirement before the mechanism\n\nWhen reviewing a change, first ask whether each mechanism is required. Flag an\nunsupported invariant or disproportionate scope before treating an agent-proposed\nchoice as settled architecture. Report these separately:\n\n- a **source-contract failure**, where required behavior or security is wrong;\n- an **optional-defense failure**, where an unrequired added defense is broken;\n- **practical severity**, rated from realistic prerequisites and consequences.\n\nDo not turn “the optional defense is broken” into “the required behavior is\ninsecure.”\n\n## Parallel test execution and isolation\n\nWhole-repository gates run test files in parallel by default. Node's test\nrunner already supplies process-level concurrency when no serializing flag is\npresent, so do not add a serializing flag to the whole gate.\n\nIf a genuinely serial suite needs serialization, give it its own named serial\nscript and put a nearby written reason next to that script. Never serialize the whole\ngate for one suite. Keep coverage wrappers such as `c8` transparent to the\nunderlying runner's concurrency: they must preserve the concurrency selected by\nthe wrapped runner rather than adding their own throttle.\n\nCI may set an optional `TEST_CONCURRENCY` control to reduce CI pressure. Accept\nthe runner default when it is unset. Reject a fixed low ceiling that ignores the\ncapacity of the host running the tests.\n\nThe slowest file sets a lower bound on parallel wall time. Treat files\naround 30 seconds or slower as split candidates. Prefer fakes or event hooks\nover `setTimeout` or sleep waits so tests do not spend their time waiting on a\nclock.\n\nMake tests parallel-safe with unique temporary directories. Avoid fixed shared\npaths, socket names, ports, daemons, and other uncontrolled shared state.\nSerialization pressure should trigger an isolation review. Fix the shared\nresource rather than hiding the collision behind a serial gate.\n\n## Before you claim it works\n\nType-check against the real typings rather than trusting an event name that\nlooks plausible. An unknown event name is a silent no-op — the handler simply\nnever runs, and nothing reports it.\n",{"load":24,"name":5,"description":7},"recognition","main","a341ec69437d028e71ea1de51b7f6d9c1238048b","packages\u002Fpi-dev\u002Fskills\u002Fauthoring-pi-extensions","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI","https:\u002F\u002Fwww.centerforairesearch.org\u002F","CAAIR develops applied AI systems that support civil society, public accountability, and democratic infrastructure.","2026-11-12 02:15:25",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add CenterForAgenticAI\u002Ftools --skill authoring-pi-extensions",{"tool":39,"command":40},"bunx","bunx skills add CenterForAgenticAI\u002Ftools --skill authoring-pi-extensions",{"tool":42,"command":43},"pnpm","pnpm dlx skills add CenterForAgenticAI\u002Ftools --skill authoring-pi-extensions","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-dev\u002Fskills\u002Fauthoring-pi-extensions\u002FSKILL.md",true,[],[48,55,62,69,76,83],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M7DBS96C51VE1KHCM1N0SR47","work-execution","Execute a validated pi-work node or wave when you need the ready-set, by-reference work_plan brief, Fabric default, pi-delegate fallback, explicit tree-bound verification, remediation, scope amendments, or a post-integration refactor pass. Use after decomposition and before claiming any node is done.","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-work\u002Fskills\u002Fwork-execution\u002FSKILL.md",[],[],{"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},"01M7DBS2SJ4ZND4SG77JX66BWK","work-decomposition","Turn a validated pi-work intent into a strict workspec work graph when the work is too large for one bounded worker. Use when assigning criterion homes, dependencies, honest write scopes, open decisions, or optional model-change nodes before planning execution.","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-work\u002Fskills\u002Fwork-decomposition\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},"01M7DBRXT98J0Q84K38FCFQ9VT","work-authoring","Draft a pi-work workspec when an outcome, scope, or proof obligation is still unclear. Use this before turning a conversation, issue, or design into a markdown draft, especially when acceptance criteria need to survive lossless promotion into strict YAML.","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-work\u002Fskills\u002Fwork-authoring\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M7DBRST29F8J729DHGCM55WJ","intent-records","Draft or approve an intent record when a person asks for a change, or describe existing behaviour before retrofitting compiled intent. Use before turning intent into a pi-work workspec.","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-intent\u002Fskills\u002Fintent-records\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M7DBRNVNYBW7K25Y7W7CYVEA","intent-conformance","Connect an approved finite Bend model to actual application code with a language-neutral conformance manifest, app-language oracles and a broken-model control. Load when proving implementation parity or diagnosing decision drift.","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-intent\u002Fskills\u002Fintent-conformance\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M7DBRH9KT2DMXVT16P1JCHQK","compiled-intent-model","Author or change a finite Bend model from an approved intent record, review its meaning, obtain human law approval, and produce an offline-verifiable Jev receipt. Load for model-change nodes, not for emergent runtime behaviour.","https:\u002F\u002Fgithub.com\u002FCenterForAgenticAI\u002Ftools\u002Fblob\u002Fmain\u002Fpackages\u002Fpi-intent\u002Fskills\u002Fcompiled-intent-model\u002FSKILL.md",[],[]]