@pencil-agent/nano-pencil 2.0.0-beta.8 → 2.0.0

package/dist/extensions/builtin/grub/README.md

@@ -1,112 +1,112 @@
-# Grub Extension
-
-`/grub` runs one autonomous long-running task until the agent reports it
-complete, reports it is blocked, the user stops it, or a safety limit is
-reached. The harness design follows the pattern described in Anthropic's
-[Effective harnesses for long-running agents](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents):
-a structured on-disk state lets coding agents pick up where they left off
-even across fresh context windows and full process restarts.
-
-## Commands
-
-- `/grub <goal> [--max-iter N] [--max-fail N]` — start one autonomous task
-- `/grub status [--json]` — show the active or last finished grub task
-- `/grub resume` — resume dispatch for an adopted/persisted task
-- `/grub stop` — stop the active grub task
-
-## Harness artifacts
-
-Each task owns a directory at `.grub/<task-id>/`:
-
-| File | Purpose | Who may write |
-|------|---------|---------------|
-| `feature-list.json` | Structured list of end-to-end features | Initializer writes the whole file once; coding agents may only flip `passes` and set `evidence` |
-| `progress-log.md` | Dated notes describing each iteration | Agent, append-only |
-| `init.sh` | Get-bearings + project smoke script run at the start of every iteration | Initializer; later agents may add project-specific smoke commands |
-| `state.json` | Durable `GrubTaskState` snapshot for cross-session resume | Controller, atomic writes |
-
-`feature-list.json` schema (version 1):
-
-```json
-{
-  "version": 1,
-  "goal": "<user goal>",
-  "features": [
-    {
-      "id": "kebab-slug",
-      "category": "functional|verification|polish",
-      "description": "observable behavior",
-      "steps": ["actionable", "verification", "steps"],
-      "passes": false,
-      "evidence": "optional git sha or short proof"
-    }
-  ]
-}
-```
-
-The controller validates every mutation: changing `description`, `steps`,
-`category`, `id`, list length, or reordering counts as a violation. Agents
-are told up front that the only permitted edits are toggling `passes` and
-setting `evidence`.
-
-## How it works
-
-- Each grub iteration is tagged with a `[GRUB:<id>:<n>]` prompt prefix so
-  the extension can recognise its own injected turns.
-- On start, grub creates the harness directory and writes the initial
-  artifacts without creating a git commit. The harness keeps durable state on
-  disk and leaves source changes visible in the working tree, avoiding noisy
-  `grub(...)` commits unless the user explicitly asks for them.
-- Two phase-specialized system prompts are injected via
-  `before_agent_start`:
-  - **Initializer prompt** (first successful turn): expand
-    `feature-list.json` into 15-40 concrete features, harden `init.sh`,
-    seed `progress-log.md`. No broad implementation yet.
-  - **Coding prompt** (remaining turns): run `init.sh`, pick exactly one
-    pending feature, implement + verify end-to-end, flip `passes` +
-    `evidence`, and append to `progress-log.md`.
-- At the end of every grub turn the assistant must emit a single
-  `<loop-state>{"status":"continue|complete|blocked", "summary":"...", "nextStep":"..."}</loop-state>`
-  block. The extension parses it and dispatches the next iteration or
-  stops with a terminal status.
-- **Initializer sanitize-not-fail**: while in the initializer phase, only
-  genuinely unfixable structural problems (feature count out of 15-40,
-  unreplaced placeholder, non-kebab ids, duplicate ids) fail the turn and
-  force a retry. Recoverable hygiene issues are auto-corrected instead of
-  killing the task: a wrong `goal` is restored to the authoritative task goal,
-  pre-marked `passes:true` are reset to `false`, and stray `evidence` is
-  dropped. The sanitized list is written back to disk and becomes the baseline,
-  so the phase always advances once the structure is valid.
-- **Phase-aware failure budget**: the initializer gets a more forgiving budget
-  (default 5) than execution (default 3, via `--max-fail`), because standing up
-  a valid harness is a distinct, retry-friendly activity from execution work.
-- **Mutation guard**: after the initializer creates the first real
-  `feature-list.json`, each subsequent turn is diffed against the persisted
-  baseline. Rewriting feature ids, descriptions, categories, steps, count, or
-  order is rejected and retried; only `passes` and `evidence` may change.
-- **Completion guard**: if the decision says `complete` but
-  `feature-list.json` still has `passes:false` entries, the controller
-  rewrites the decision to `continue` with a synthetic `nextStep`
-  pointing at the first pending feature. The harness will not allow
-  premature "done".
-- **Cross-session resume**: `GrubTaskState` is written atomically to
-  `state.json` on every transition. On the next session, `session_start`
-  calls `discoverActiveTasks()` and adopts the most recent running task
-  without auto-dispatching — the user types `/grub resume` to continue.
-- **Safety limits**: 25 iterations and 3 consecutive failures by default;
-  override with `--max-iter` / `--max-fail`.
-- **Stale harness cleanup**: on extension load, terminal harnesses older
-  than 30 days are pruned from `.grub/`.
-
-## Legacy migration
-
-Earlier versions wrote `feature-checklist.md` (markdown checkboxes). When a
-new iteration starts and `feature-list.json` is missing but the legacy file
-exists, its checkbox items are migrated into the JSON format (category
-defaults to `functional`; `steps` start empty so the initializer can refine
-later).
-
-## Related
-
-For the recurring scheduler that runs prompts or slash commands on an
-interval, see the sibling [`loop` extension](../loop/README.md).
+# Grub Extension
+
+`/grub` runs one autonomous long-running task until the agent reports it
+complete, reports it is blocked, the user stops it, or a safety limit is
+reached. The harness design follows the pattern described in Anthropic's
+[Effective harnesses for long-running agents](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents):
+a structured on-disk state lets coding agents pick up where they left off
+even across fresh context windows and full process restarts.
+
+## Commands
+
+- `/grub <goal> [--max-iter N] [--max-fail N]` — start one autonomous task
+- `/grub status [--json]` — show the active or last finished grub task
+- `/grub resume` — resume dispatch for an adopted/persisted task
+- `/grub stop` — stop the active grub task
+
+## Harness artifacts
+
+Each task owns a directory at `.grub/<task-id>/`:
+
+| File | Purpose | Who may write |
+|------|---------|---------------|
+| `feature-list.json` | Structured list of end-to-end features | Initializer writes the whole file once; coding agents may only flip `passes` and set `evidence` |
+| `progress-log.md` | Dated notes describing each iteration | Agent, append-only |
+| `init.sh` | Get-bearings + project smoke script run at the start of every iteration | Initializer; later agents may add project-specific smoke commands |
+| `state.json` | Durable `GrubTaskState` snapshot for cross-session resume | Controller, atomic writes |
+
+`feature-list.json` schema (version 1):
+
+```json
+{
+  "version": 1,
+  "goal": "<user goal>",
+  "features": [
+    {
+      "id": "kebab-slug",
+      "category": "functional|verification|polish",
+      "description": "observable behavior",
+      "steps": ["actionable", "verification", "steps"],
+      "passes": false,
+      "evidence": "optional git sha or short proof"
+    }
+  ]
+}
+```
+
+The controller validates every mutation: changing `description`, `steps`,
+`category`, `id`, list length, or reordering counts as a violation. Agents
+are told up front that the only permitted edits are toggling `passes` and
+setting `evidence`.
+
+## How it works
+
+- Each grub iteration is tagged with a `[GRUB:<id>:<n>]` prompt prefix so
+  the extension can recognise its own injected turns.
+- On start, grub creates the harness directory and writes the initial
+  artifacts without creating a git commit. The harness keeps durable state on
+  disk and leaves source changes visible in the working tree, avoiding noisy
+  `grub(...)` commits unless the user explicitly asks for them.
+- Two phase-specialized system prompts are injected via
+  `before_agent_start`:
+  - **Initializer prompt** (first successful turn): expand
+    `feature-list.json` into 15-40 concrete features, harden `init.sh`,
+    seed `progress-log.md`. No broad implementation yet.
+  - **Coding prompt** (remaining turns): run `init.sh`, pick exactly one
+    pending feature, implement + verify end-to-end, flip `passes` +
+    `evidence`, and append to `progress-log.md`.
+- At the end of every grub turn the assistant must emit a single
+  `<loop-state>{"status":"continue|complete|blocked", "summary":"...", "nextStep":"..."}</loop-state>`
+  block. The extension parses it and dispatches the next iteration or
+  stops with a terminal status.
+- **Initializer sanitize-not-fail**: while in the initializer phase, only
+  genuinely unfixable structural problems (feature count out of 15-40,
+  unreplaced placeholder, non-kebab ids, duplicate ids) fail the turn and
+  force a retry. Recoverable hygiene issues are auto-corrected instead of
+  killing the task: a wrong `goal` is restored to the authoritative task goal,
+  pre-marked `passes:true` are reset to `false`, and stray `evidence` is
+  dropped. The sanitized list is written back to disk and becomes the baseline,
+  so the phase always advances once the structure is valid.
+- **Phase-aware failure budget**: the initializer gets a more forgiving budget
+  (default 5) than execution (default 3, via `--max-fail`), because standing up
+  a valid harness is a distinct, retry-friendly activity from execution work.
+- **Mutation guard**: after the initializer creates the first real
+  `feature-list.json`, each subsequent turn is diffed against the persisted
+  baseline. Rewriting feature ids, descriptions, categories, steps, count, or
+  order is rejected and retried; only `passes` and `evidence` may change.
+- **Completion guard**: if the decision says `complete` but
+  `feature-list.json` still has `passes:false` entries, the controller
+  rewrites the decision to `continue` with a synthetic `nextStep`
+  pointing at the first pending feature. The harness will not allow
+  premature "done".
+- **Cross-session resume**: `GrubTaskState` is written atomically to
+  `state.json` on every transition. On the next session, `session_start`
+  calls `discoverActiveTasks()` and adopts the most recent running task
+  without auto-dispatching — the user types `/grub resume` to continue.
+- **Safety limits**: 25 iterations and 3 consecutive failures by default;
+  override with `--max-iter` / `--max-fail`.
+- **Stale harness cleanup**: on extension load, terminal harnesses older
+  than 30 days are pruned from `.grub/`.
+
+## Legacy migration
+
+Earlier versions wrote `feature-checklist.md` (markdown checkboxes). When a
+new iteration starts and `feature-list.json` is missing but the legacy file
+exists, its checkbox items are migrated into the JSON format (category
+defaults to `functional`; `steps` start empty so the initializer can refine
+later).
+
+## Related
+
+For the recurring scheduler that runs prompts or slash commands on an
+interval, see the sibling [`loop` extension](../loop/README.md).

package/dist/extensions/builtin/link-world/agent-workspace/README.md

@@ -1,16 +1,16 @@
-# Link-world Workspace
-
-This directory is copied to:
-
-```text
-.nanopencil/link-world-workspace/
-```
-
-Use it to store project-local, reusable internet-access knowledge for NanoPencil.
-
-Suggested layout:
-
-- `domain-skills/<site>/...` for durable site-specific notes
-- `notes/` for broader task or provider observations that should not live in the shipped extension
-
-Keep public-safe content only. Do not store secrets, cookies, tokens, or user-specific private data here.
+# Link-world Workspace
+
+This directory is copied to:
+
+```text
+.nanopencil/link-world-workspace/
+```
+
+Use it to store project-local, reusable internet-access knowledge for NanoPencil.
+
+Suggested layout:
+
+- `domain-skills/<site>/...` for durable site-specific notes
+- `notes/` for broader task or provider observations that should not live in the shipped extension
+
+Keep public-safe content only. Do not store secrets, cookies, tokens, or user-specific private data here.

package/dist/extensions/builtin/link-world/index.js

@@ -1,19 +1,19 @@
-var O=Object.defineProperty;var a=(t,e)=>O(t,"name",{value:e,configurable:!0});import{execFile as R,execSync as T}from"node:child_process";import{cpSync as M,existsSync as g,mkdirSync as v,readFileSync as W,readdirSync as N,statSync as j}from"node:fs";import{dirname as D,join as f,resolve as H}from"node:path";import{fileURLToPath as F}from"node:url";import{Type as i}from"@sinclair/typebox";import{Box as q,Container as B,Spacer as C,Text as K}from"@pencil-agent/tui";import{getLinkWorldWorkspaceDir as J}from"../../../config.js";const m=D(F(import.meta.url)),L=f(m,"linkworld.md"),A=f(m,"internet-search","internet-search.md"),E=f(m,"link-world-agent.md"),G=f(m,"network-routing.md"),P=f(m,"agent-workspace"),h="link-world-install",w=120,Y=[{value:"status",label:"status",description:"Show installed internet-access capabilities"},{value:"doctor",label:"doctor",description:"Run agent-reach doctor diagnostics"},{value:"version",label:"version",description:"Show installed agent-reach version"},{value:"install",label:"install",description:"Show bundled installation guidance"},{value:"workspace",label:"workspace",description:"Show the link-world workspace path"},{value:"help",label:"help",description:"Show link-world commands"}],z=i.Object({action:i.Union([i.Literal("status"),i.Literal("doctor"),i.Literal("version"),i.Literal("install_help")]),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1}),V=i.Object({args:i.Array(i.String(),{description:'Arguments passed to the external `agent-reach` CLI, for example ["search", "OpenAI Responses API"] or ["doctor"].'}),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1}),Q=i.Object({query:i.String({description:"Search query to send through link-world."}),provider:i.Optional(i.String({description:"Optional provider or site hint, if the downstream runtime supports it."})),limit:i.Optional(i.Number({description:"Optional result count hint, if the downstream runtime supports it."})),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1}),X=i.Object({url:i.String({description:"URL to fetch through link-world."}),provider:i.Optional(i.String({description:"Optional provider or site hint, if the downstream runtime supports it."})),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1});function Z(){return J()}a(Z,"linkWorldWorkspacePath");function y(){const t=Z();return g(t)||(v(t,{recursive:!0}),g(P)?M(P,t,{recursive:!0}):v(t,{recursive:!0})),t}a(y,"ensureLinkWorldWorkspace");function I(t){if(!g(t))return[];const e=[];for(const n of N(t)){const o=f(t,n);if(j(o).isDirectory()){e.push(...I(o));continue}n.endsWith(".md")&&e.push(o)}return e.sort()}a(I,"collectMarkdownFiles");function ee(){try{return W(L,"utf-8")}catch{return""}}a(ee,"getInstallDoc");function p(){try{return T("agent-reach --version",{encoding:"utf-8",stdio:"pipe"}),!0}catch{return!1}}a(p,"isAgentReachInstalled");function te(){try{return T("agent-reach --help",{encoding:"utf-8",stdio:"pipe"})}catch{return""}}a(te,"getAgentReachHelpText");function _(){if(!p())return{search:!1,fetch:!1};const t=te().toLowerCase();return{search:/\bsearch\b/.test(t),fetch:/\bfetch\b/.test(t)}}a(_,"getLinkWorldCapabilities");function k(t,e,n){return new Promise((o,r)=>{R("agent-reach",t,{timeout:e*1e3,windowsHide:!0,signal:n,maxBuffer:1024*1024,encoding:"utf-8"},(s,c,d)=>{const l=[c,d].filter(Boolean).join("").trim()||"(no output)";if(s){const u=s;if(u.code==="ENOENT"){r(new Error("agent-reach is not installed"));return}o({exitCode:typeof u.code=="number"?u.code:1,output:l});return}o({exitCode:0,output:l})})})}a(k,"execAgentReach");const ne="https://r.jina.ai",re="https://s.jina.ai",$=3e4;async function oe(t,e){const n=t.startsWith("http")?t:`https://${t}`,o=e??AbortSignal.timeout($);try{const r=await fetch(`${ne}/${n}`,{signal:o,headers:{Accept:"text/markdown"}});if(r.ok){const s=await r.text();if(s.trim())return s}}catch{}try{const r=await fetch(n,{signal:o});if(!r.ok)throw new Error(`HTTP ${r.status}: ${r.statusText}`);const s=await r.text();return ie(s,n)}catch(r){throw new Error(`Failed to fetch ${n}: ${r instanceof Error?r.message:r}`)}}a(oe,"nativeWebFetch");async function se(t,e,n){const o=n??AbortSignal.timeout($);try{const r=new URLSearchParams({q:t});e>0&&r.set("num",String(Math.min(e,10)));const s=await fetch(`${re}?${r}`,{signal:o,headers:{Accept:"text/markdown"}});if(s.ok){const c=await s.text();if(c.trim())return c}}catch{}try{const r=new URLSearchParams({q:t}),s=await fetch(`https://html.duckduckgo.com/html/?${r}`,{signal:o,headers:{"User-Agent":"Mozilla/5.0 (compatible; NanoPencil/1.0)"}});if(s.ok){const c=await s.text(),d=ae(c,e||5);if(d.length>0)return d}}catch{}try{const r=new URLSearchParams({q:t,format:"json",no_html:"1"}),s=await fetch(`https://api.duckduckgo.com/?${r}`,{signal:o});if(s.ok){const c=await s.json(),d=typeof c.AbstractText=="string"?c.AbstractText:"",l=typeof c.AbstractSource=="string"?c.AbstractSource:"",u=typeof c.AbstractURL=="string"?c.AbstractURL:"";if(d)return[`## ${t}`,"",d,l?`Source: ${l}`:"",u?`URL: ${u}`:""].filter(Boolean).join(`
+var U=Object.defineProperty;var a=(t,e)=>U(t,"name",{value:e,configurable:!0});import{execFile as O,execSync as T}from"node:child_process";import{cpSync as R,existsSync as g,mkdirSync as v,readFileSync as M,readdirSync as W,statSync as N}from"node:fs";import{dirname as j,join as f,resolve as D}from"node:path";import{fileURLToPath as H}from"node:url";import{Type as i}from"@sinclair/typebox";import{Box as F,Container as q,Spacer as C,Text as B}from"@pencil-agent/tui";import{getLinkWorldWorkspaceDir as K}from"../../../config.js";const m=j(H(import.meta.url)),J=f(m,"linkworld.md"),L=f(m,"internet-search","internet-search.md"),A=f(m,"link-world-agent.md"),G=f(m,"network-routing","network-routing.md"),E=f(m,"agent-workspace"),h="link-world-install",w=120,Y=[{value:"status",label:"status",description:"Show installed internet-access capabilities"},{value:"doctor",label:"doctor",description:"Run agent-reach doctor diagnostics"},{value:"version",label:"version",description:"Show installed agent-reach version"},{value:"install",label:"install",description:"Show bundled installation guidance"},{value:"workspace",label:"workspace",description:"Show the link-world workspace path"},{value:"help",label:"help",description:"Show link-world commands"}],z=i.Object({action:i.Union([i.Literal("status"),i.Literal("doctor"),i.Literal("version"),i.Literal("install_help")]),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1}),V=i.Object({args:i.Array(i.String(),{description:'Arguments passed to the external `agent-reach` CLI, for example ["search", "OpenAI Responses API"] or ["doctor"].'}),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1}),Q=i.Object({query:i.String({description:"Search query to send through link-world."}),provider:i.Optional(i.String({description:"Optional provider or site hint, if the downstream runtime supports it."})),limit:i.Optional(i.Number({description:"Optional result count hint, if the downstream runtime supports it."})),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1}),X=i.Object({url:i.String({description:"URL to fetch through link-world."}),provider:i.Optional(i.String({description:"Optional provider or site hint, if the downstream runtime supports it."})),timeout:i.Optional(i.Number({description:"Timeout in seconds. Defaults to 120."}))},{additionalProperties:!1});function Z(){return K()}a(Z,"linkWorldWorkspacePath");function y(){const t=Z();return g(t)||(v(t,{recursive:!0}),g(E)?R(E,t,{recursive:!0}):v(t,{recursive:!0})),t}a(y,"ensureLinkWorldWorkspace");function P(t){if(!g(t))return[];const e=[];for(const n of W(t)){const o=f(t,n);if(N(o).isDirectory()){e.push(...P(o));continue}n.endsWith(".md")&&e.push(o)}return e.sort()}a(P,"collectMarkdownFiles");function ee(){try{return M(J,"utf-8")}catch{return""}}a(ee,"getInstallDoc");function p(){try{return T("agent-reach --version",{encoding:"utf-8",stdio:"pipe"}),!0}catch{return!1}}a(p,"isAgentReachInstalled");function te(){try{return T("agent-reach --help",{encoding:"utf-8",stdio:"pipe"})}catch{return""}}a(te,"getAgentReachHelpText");function _(){if(!p())return{search:!1,fetch:!1};const t=te().toLowerCase();return{search:/\bsearch\b/.test(t),fetch:/\bfetch\b/.test(t)}}a(_,"getLinkWorldCapabilities");function k(t,e,n){return new Promise((o,r)=>{O("agent-reach",t,{timeout:e*1e3,windowsHide:!0,signal:n,maxBuffer:1024*1024,encoding:"utf-8"},(s,c,d)=>{const l=[c,d].filter(Boolean).join("").trim()||"(no output)";if(s){const u=s;if(u.code==="ENOENT"){r(new Error("agent-reach is not installed"));return}o({exitCode:typeof u.code=="number"?u.code:1,output:l});return}o({exitCode:0,output:l})})})}a(k,"execAgentReach");const ne="https://r.jina.ai",re="https://s.jina.ai",I=3e4;async function oe(t,e){const n=t.startsWith("http")?t:`https://${t}`,o=e??AbortSignal.timeout(I);try{const r=await fetch(`${ne}/${n}`,{signal:o,headers:{Accept:"text/markdown"}});if(r.ok){const s=await r.text();if(s.trim())return s}}catch{}try{const r=await fetch(n,{signal:o});if(!r.ok)throw new Error(`HTTP ${r.status}: ${r.statusText}`);const s=await r.text();return ie(s,n)}catch(r){throw new Error(`Failed to fetch ${n}: ${r instanceof Error?r.message:r}`)}}a(oe,"nativeWebFetch");async function se(t,e,n){const o=n??AbortSignal.timeout(I);try{const r=new URLSearchParams({q:t});e>0&&r.set("num",String(Math.min(e,10)));const s=await fetch(`${re}?${r}`,{signal:o,headers:{Accept:"text/markdown"}});if(s.ok){const c=await s.text();if(c.trim())return c}}catch{}try{const r=new URLSearchParams({q:t}),s=await fetch(`https://html.duckduckgo.com/html/?${r}`,{signal:o,headers:{"User-Agent":"Mozilla/5.0 (compatible; NanoPencil/1.0)"}});if(s.ok){const c=await s.text(),d=ae(c,e||5);if(d.length>0)return d}}catch{}try{const r=new URLSearchParams({q:t,format:"json",no_html:"1"}),s=await fetch(`https://api.duckduckgo.com/?${r}`,{signal:o});if(s.ok){const c=await s.json(),d=typeof c.AbstractText=="string"?c.AbstractText:"",l=typeof c.AbstractSource=="string"?c.AbstractSource:"",u=typeof c.AbstractURL=="string"?c.AbstractURL:"";if(d)return[`## ${t}`,"",d,l?`Source: ${l}`:"",u?`URL: ${u}`:""].filter(Boolean).join(`
 `)}}catch{}throw new Error(`Search failed for "${t}": all providers returned errors. Check your network connection.`)}a(se,"nativeWebSearch");function ie(t,e){let n=t;return n=n.replace(/<(script|style|nav|header|footer|noscript)[^>]*>[\s\S]*?<\/\1>/gi,""),n=n.replace(/<[^>]+>/g," "),n=n.replace(/&amp;/g,"&").replace(/&lt;/g,"<").replace(/&gt;/g,">").replace(/&quot;/g,'"').replace(/&#39;/g,"'").replace(/&nbsp;/g," "),n=n.replace(/[ \t]+/g," ").replace(/\n{3,}/g,`
 
 `).trim(),`# Content from ${e}
 
-${n}`}a(ie,"htmlToPlainText");function ae(t,e){const n=[],o=/<a[^>]+class="result__a"[^>]*href="([^"]*)"[^>]*>([\s\S]*?)<\/a>/gi,r=/<a[^>]+class="result__snippet"[^>]*>([\s\S]*?)<\/a>/gi,s=[],c=[];let d;for(;(d=o.exec(t))!==null;){const l=d[1],u=d[2].replace(/<[^>]+>/g,"").trim(),S=l.match(/uddg=([^&]+)/),U=S?decodeURIComponent(S[1]):l;u&&s.push({url:U,title:u})}for(;(d=r.exec(t))!==null;){const l=d[1].replace(/<[^>]+>/g,"").trim();l&&c.push(l)}for(let l=0;l<Math.min(s.length,e);l++){const u=[`### ${s[l].title}`,s[l].url];c[l]&&u.push(c[l]),n.push(u.join(`
+${n}`}a(ie,"htmlToPlainText");function ae(t,e){const n=[],o=/<a[^>]+class="result__a"[^>]*href="([^"]*)"[^>]*>([\s\S]*?)<\/a>/gi,r=/<a[^>]+class="result__snippet"[^>]*>([\s\S]*?)<\/a>/gi,s=[],c=[];let d;for(;(d=o.exec(t))!==null;){const l=d[1],u=d[2].replace(/<[^>]+>/g,"").trim(),S=l.match(/uddg=([^&]+)/),$=S?decodeURIComponent(S[1]):l;u&&s.push({url:$,title:u})}for(;(d=r.exec(t))!==null;){const l=d[1].replace(/<[^>]+>/g,"").trim();l&&c.push(l)}for(let l=0;l<Math.min(s.length,e);l++){const u=[`### ${s[l].title}`,s[l].url];c[l]&&u.push(c[l]),n.push(u.join(`
 `))}return n.length>0?n.join(`
 
 `):""}a(ae,"parseDuckDuckGoResults");function x(){const t=ee();return t?["link-world is not bundled yet, so installation still targets the external `agent-reach` runtime.","","Use this document as the source of truth:","",t].join(`
 `):["link-world is not bundled yet, and the local install guide is missing.","Use the official install document:","https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md"].join(`
-`)}a(x,"installHelpText");function b(){const t=_(),e=p(),n=["Link-world status","",`agent-reach installed: ${e?"yes":"no (native fallback active)"}`,`internet-search skill bundled: ${g(A)?"yes":"no"}`,`agent guidance bundled: ${g(E)?"yes":"no"}`,`web_search: ${t.search?"agent-reach":"native (Jina Search)"}`,`web_fetch: ${t.fetch?"agent-reach":"native (Jina Reader)"}`];return e||(n.push(""),n.push("Native fallback provides basic search and page fetching without agent-reach."),n.push("For advanced platform channels (Twitter, YouTube, etc.), run `/link-world install`.")),n.join(`
+`)}a(x,"installHelpText");function b(){const t=_(),e=p(),n=["Link-world status","",`agent-reach installed: ${e?"yes":"no (native fallback active)"}`,`internet-search skill bundled: ${g(L)?"yes":"no"}`,`agent guidance bundled: ${g(A)?"yes":"no"}`,`web_search: ${t.search?"agent-reach":"native (Jina Search)"}`,`web_fetch: ${t.fetch?"agent-reach":"native (Jina Reader)"}`];return e||(n.push(""),n.push("Native fallback provides basic search and page fetching without agent-reach."),n.push("For advanced platform channels (Twitter, YouTube, etc.), run `/link-world install`.")),n.join(`
 `)}a(b,"getStatusText");function ce(){return{name:"link_world_admin",label:"link-world admin",description:"Inspect or troubleshoot the built-in link-world integration. Use for install guidance, version checks, and agent-reach doctor/status commands.",parameters:z,guidance:"Use this before attempting internet tasks if you are unsure whether agent-reach is installed. Prefer action=status or action=doctor for diagnostics.",execute:a(async(t,e,n)=>{if(e.action==="install_help")return{content:[{type:"text",text:x()}],details:null};if(e.action==="status")return{content:[{type:"text",text:b()}],details:null};if(!p())throw new Error(`${b()}
 
 ${x()}`);const o=e.action==="doctor"?["doctor"]:["--version"],r=await k(o,e.timeout??w,n);if(r.exitCode&&r.exitCode!==0)throw new Error(r.output);return{content:[{type:"text",text:r.output}],details:{exitCode:r.exitCode}}},"execute")}}a(ce,"createLinkWorldAdminTool");function le(){return{name:"link_world_exec",label:"link-world exec",description:"Execute the external `agent-reach` CLI through NanoPencil. Use this for actual internet tasks once link-world is installed, instead of going through the bash tool.",parameters:V,guidance:"Use `link_world_admin` first if you are unsure whether agent-reach is installed. Pass explicit CLI arguments, not a shell string. Prefer site or domain skills before inventing new agent-reach commands.",execute:a(async(t,e,n)=>{if(!p())throw new Error(`${b()}
 
-${x()}`);if(e.args.length===0)throw new Error("link_world_exec requires at least one CLI argument");const o=await k(e.args,e.timeout??w,n);if(o.exitCode&&o.exitCode!==0)throw new Error(o.output);return{content:[{type:"text",text:o.output}],details:{exitCode:o.exitCode,args:e.args}}},"execute")}}a(le,"createLinkWorldExecTool");function de(){return{name:"web_search",label:"web search",description:"High-level internet search through link-world. Prefer this for normal search tasks instead of constructing raw agent-reach argv yourself.",parameters:Q,guidance:"Use this for ordinary web search and lightweight research tasks. Use link_world_exec only when you need a lower-level agent-reach command that this tool does not model.",execute:a(async(t,e,n)=>{if(p()&&_().search){const s=["search",e.query];e.provider&&s.push("--provider",e.provider),typeof e.limit=="number"&&Number.isFinite(e.limit)&&s.push("--limit",String(e.limit));const c=await k(s,e.timeout??w,n);if(c.exitCode&&c.exitCode!==0)throw new Error(c.output);return{content:[{type:"text",text:c.output}],details:{exitCode:c.exitCode,args:s}}}const o=typeof e.limit=="number"&&Number.isFinite(e.limit)?e.limit:5;return{content:[{type:"text",text:await se(e.query,o,n)}],details:{fallback:"jina",query:e.query}}},"execute")}}a(de,"createWebSearchTool");function ue(){return{name:"web_fetch",label:"web fetch",description:"High-level URL fetch through link-world. Prefer this when the user already has a target page or endpoint and needs its content.",parameters:X,guidance:"Use this when the user provides a URL or when a prior search result should be fetched directly. If the task requires page interaction, use the browser tool family instead.",execute:a(async(t,e,n)=>{if(p()&&_().fetch){const r=["fetch",e.url];e.provider&&r.push("--provider",e.provider);const s=await k(r,e.timeout??w,n);if(s.exitCode&&s.exitCode!==0)throw new Error(s.output);return{content:[{type:"text",text:s.output}],details:{exitCode:s.exitCode,args:r}}}return{content:[{type:"text",text:await oe(e.url,n)}],details:{fallback:"jina",url:e.url}}},"execute")}}a(ue,"createWebFetchTool");function he(t){t.registerMessageRenderer(h,(e,n,o)=>{const r=new q(1,1,l=>o.bg("customMessageBg",l)),s=o.fg("customMessageLabel","\x1B[1m[link-world]\x1B[22m "),c=o.fg("customMessageText",String(e.content??""));r.addChild(new K(s+c,0,0)),r.addChild(new C(1));const d=new B;return d.addChild(new C(1)),d.addChild(r),d}),t.registerTool(ce()),t.registerTool(le()),t.registerTool(de()),t.registerTool(ue()),t.on("session_start",(e,n)=>{y()}),t.registerCommand("link-world",{description:"Set up or inspect internet access tools",getArgumentCompletions:a((e,n)=>{if(n&&n.tokenIndex>0)return null;const o=e.trim().toLowerCase(),r=Y.filter(s=>s.value.startsWith(o));return r.length>0?r:null},"getArgumentCompletions"),handler:a(async(e,n)=>{const o=(e.trim().split(/\s+/)[0]||"help").toLowerCase();if(o==="help"){const r=p()?"agent-reach":"native (Jina)";t.sendMessage({customType:h,content:["Link-world is NanoPencil's built-in internet access integration point.","",`Mode: ${r}`,"","Commands:","/link-world status  - show whether agent-reach is installed and whether skills are bundled","/link-world doctor  - run `agent-reach doctor`","/link-world version - print the installed agent-reach version","/link-world install - show the bundled installation guide","/link-world workspace - show the project-local link-world workspace","","Tools: `web_search`, `web_fetch`, `link_world_admin`, `link_world_exec`"].join(`
-`),display:!0});return}if(o==="workspace"){t.sendMessage({customType:h,content:`Link-world workspace: ${H(y())}`,display:!0});return}if(o==="install"){t.sendMessage({customType:h,content:x(),display:!0});return}if(o==="status"){t.sendMessage({customType:h,content:b(),display:!0});return}if(o==="doctor"||o==="version"){if(!p()){t.sendMessage({customType:h,content:`${b()}
+${x()}`);if(e.args.length===0)throw new Error("link_world_exec requires at least one CLI argument");const o=await k(e.args,e.timeout??w,n);if(o.exitCode&&o.exitCode!==0)throw new Error(o.output);return{content:[{type:"text",text:o.output}],details:{exitCode:o.exitCode,args:e.args}}},"execute")}}a(le,"createLinkWorldExecTool");function de(){return{name:"web_search",label:"web search",description:"High-level internet search through link-world. Prefer this for normal search tasks instead of constructing raw agent-reach argv yourself.",parameters:Q,guidance:"Use this for ordinary web search and lightweight research tasks. Use link_world_exec only when you need a lower-level agent-reach command that this tool does not model.",execute:a(async(t,e,n)=>{if(p()&&_().search){const s=["search",e.query];e.provider&&s.push("--provider",e.provider),typeof e.limit=="number"&&Number.isFinite(e.limit)&&s.push("--limit",String(e.limit));const c=await k(s,e.timeout??w,n);if(c.exitCode&&c.exitCode!==0)throw new Error(c.output);return{content:[{type:"text",text:c.output}],details:{exitCode:c.exitCode,args:s}}}const o=typeof e.limit=="number"&&Number.isFinite(e.limit)?e.limit:5;return{content:[{type:"text",text:await se(e.query,o,n)}],details:{fallback:"jina",query:e.query}}},"execute")}}a(de,"createWebSearchTool");function ue(){return{name:"web_fetch",label:"web fetch",description:"High-level URL fetch through link-world. Prefer this when the user already has a target page or endpoint and needs its content.",parameters:X,guidance:"Use this when the user provides a URL or when a prior search result should be fetched directly. If the task requires page interaction, use the browser tool family instead.",execute:a(async(t,e,n)=>{if(p()&&_().fetch){const r=["fetch",e.url];e.provider&&r.push("--provider",e.provider);const s=await k(r,e.timeout??w,n);if(s.exitCode&&s.exitCode!==0)throw new Error(s.output);return{content:[{type:"text",text:s.output}],details:{exitCode:s.exitCode,args:r}}}return{content:[{type:"text",text:await oe(e.url,n)}],details:{fallback:"jina",url:e.url}}},"execute")}}a(ue,"createWebFetchTool");function he(t){t.registerMessageRenderer(h,(e,n,o)=>{const r=new F(1,1,l=>o.bg("customMessageBg",l)),s=o.fg("customMessageLabel","\x1B[1m[link-world]\x1B[22m "),c=o.fg("customMessageText",String(e.content??""));r.addChild(new B(s+c,0,0)),r.addChild(new C(1));const d=new q;return d.addChild(new C(1)),d.addChild(r),d}),t.registerTool(ce()),t.registerTool(le()),t.registerTool(de()),t.registerTool(ue()),t.on("session_start",(e,n)=>{y()}),t.registerCommand("link-world",{description:"Set up or inspect internet access tools",getArgumentCompletions:a((e,n)=>{if(n&&n.tokenIndex>0)return null;const o=e.trim().toLowerCase(),r=Y.filter(s=>s.value.startsWith(o));return r.length>0?r:null},"getArgumentCompletions"),handler:a(async(e,n)=>{const o=(e.trim().split(/\s+/)[0]||"help").toLowerCase();if(o==="help"){const r=p()?"agent-reach":"native (Jina)";t.sendMessage({customType:h,content:["Link-world is NanoPencil's built-in internet access integration point.","",`Mode: ${r}`,"","Commands:","/link-world status  - show whether agent-reach is installed and whether skills are bundled","/link-world doctor  - run `agent-reach doctor`","/link-world version - print the installed agent-reach version","/link-world install - show the bundled installation guide","/link-world workspace - show the project-local link-world workspace","","Tools: `web_search`, `web_fetch`, `link_world_admin`, `link_world_exec`"].join(`
+`),display:!0});return}if(o==="workspace"){t.sendMessage({customType:h,content:`Link-world workspace: ${D(y())}`,display:!0});return}if(o==="install"){t.sendMessage({customType:h,content:x(),display:!0});return}if(o==="status"){t.sendMessage({customType:h,content:b(),display:!0});return}if(o==="doctor"||o==="version"){if(!p()){t.sendMessage({customType:h,content:`${b()}
 
-${x()}`,display:!0});return}const r=await k(o==="doctor"?["doctor"]:["--version"],w);t.sendMessage({customType:h,content:r.output,display:!0});return}t.sendMessage({customType:h,content:"Usage: /link-world [help|status|doctor|version|install|workspace]",display:!0})},"handler")}),t.on("resources_discover",async e=>{const n=y(),o=[...new Set([G,E,A,L,...I(f(n,"domain-skills"))].filter(r=>g(r)))];return o.length?{skillPaths:o}:{}})}a(he,"linkWorldExtension");export{he as default};
+${x()}`,display:!0});return}const r=await k(o==="doctor"?["doctor"]:["--version"],w);t.sendMessage({customType:h,content:r.output,display:!0});return}t.sendMessage({customType:h,content:"Usage: /link-world [help|status|doctor|version|install|workspace]",display:!0})},"handler")}),t.on("resources_discover",async e=>{const n=y(),o=[...new Set([G,A,L,...P(f(n,"domain-skills"))].filter(r=>g(r)))];return o.length?{skillPaths:o}:{}})}a(he,"linkWorldExtension");export{he as default};

package/dist/extensions/builtin/link-world/internet-search/internet-search.md

@@ -1,65 +1,65 @@
----
-name: internet-search
-description: 当用户需要联网搜索最新信息、新闻、在线内容时使用此技能。包括搜索 Twitter、YouTube、Bilibili、知乎、微博、小红书等平台。
----
-
-# 互联网搜索
-
-## 触发条件
-
-当用户请求以下内容时使用此技能：
-- 搜索互联网、联网搜索、网络搜索
-- 查找最新新闻、资讯、信息
-- 搜索 YouTube、Bilibili、Twitter、小红书、微博等内容
-- 查找某个话题的最新讨论
-- 获取实时信息或热点内容
-
-## 执行步骤
-
-### 步骤 1：检查 agent-reach 是否安装
-
-运行以下命令检查 agent-reach 是否可用：
-
-```bash
-agent-reach --version
-```
-
-### 步骤 2：如果未安装
-
-如果 agent-reach 命令不可用，告知用户需要先安装：
-
-> link-world (agent-reach) 未安装。请先运行 `/link-world` 安装 agent-reach 以支持联网搜索功能。
-
-### 步骤 3：如果已安装
-
-根据用户需求，使用 linkworld.md 中定义的工具进行搜索。
-
-**完整命令参考请查阅：extensions/link-world/linkworld.md**
-
-| 场景 | 推荐工具 | 示例命令 |
-|------|----------|----------|
-| 搜索 Twitter/X 推文 | xreach | `xreach search "关键词" --json` |
-| 解析 YouTube 视频 | yt-dlp | `yt-dlp --dump-json "视频URL"` |
-| 解析 Bilibili 视频 | yt-dlp | `yt-dlp --dump-json "视频URL"` |
-| 搜索 Reddit 帖子 | curl | `curl -s "https://reddit.com/r/xxx/search.json?q=关键词"` |
-| 搜索 GitHub 仓库 | gh | `gh search repos "关键词" --limit 5` |
-| 读取网页内容 | curl + Jina | `curl -s "https://r.jina.ai/网页URL"` |
-| 通用搜索引擎 | Exa (MCP) | `mcporter call 'exa.web_search_exa(query: "关键词", num_results: 5)'` |
-| 搜索小红书笔记 | 小红书 (MCP) | `mcporter call 'xiaohongshu.search_feeds(keyword: "关键词", limit: 5)'` |
-| 解析抖音视频 | 抖音 (MCP) | `mcporter call 'douyin.parse_douyin_video_info(url: "视频URL")'` |
-
-### 步骤 4：返回结果
-
-将搜索结果整理后返回给用户，包括：
-- 来源平台
-- 标题
-- 内容摘要
-- 链接
-
-## 注意事项
-
-1. 遵守平台的 API 使用规范
-2. 如需登录平台的搜索功能，确保用户已配置认证信息
-3. 如果搜索失败，尝试其他平台或告知用户可能的错误原因
-4. YouTube/Bilibili 需要提供具体的视频 URL，不能直接搜索关键词
-5. 详细的配置说明（Cookie、代理等）请参考 linkworld.md
+---
+name: internet-search
+description: 当用户需要联网搜索最新信息、新闻、在线内容时使用此技能。包括搜索 Twitter、YouTube、Bilibili、知乎、微博、小红书等平台。
+---
+
+# 互联网搜索
+
+## 触发条件
+
+当用户请求以下内容时使用此技能：
+- 搜索互联网、联网搜索、网络搜索
+- 查找最新新闻、资讯、信息
+- 搜索 YouTube、Bilibili、Twitter、小红书、微博等内容
+- 查找某个话题的最新讨论
+- 获取实时信息或热点内容
+
+## 执行步骤
+
+### 步骤 1：检查 agent-reach 是否安装
+
+运行以下命令检查 agent-reach 是否可用：
+
+```bash
+agent-reach --version
+```
+
+### 步骤 2：如果未安装
+
+如果 agent-reach 命令不可用，告知用户需要先安装：
+
+> link-world (agent-reach) 未安装。请先运行 `/link-world` 安装 agent-reach 以支持联网搜索功能。
+
+### 步骤 3：如果已安装
+
+根据用户需求，使用 linkworld.md 中定义的工具进行搜索。
+
+**完整命令参考请查阅：extensions/link-world/linkworld.md**
+
+| 场景 | 推荐工具 | 示例命令 |
+|------|----------|----------|
+| 搜索 Twitter/X 推文 | xreach | `xreach search "关键词" --json` |
+| 解析 YouTube 视频 | yt-dlp | `yt-dlp --dump-json "视频URL"` |
+| 解析 Bilibili 视频 | yt-dlp | `yt-dlp --dump-json "视频URL"` |
+| 搜索 Reddit 帖子 | curl | `curl -s "https://reddit.com/r/xxx/search.json?q=关键词"` |
+| 搜索 GitHub 仓库 | gh | `gh search repos "关键词" --limit 5` |
+| 读取网页内容 | curl + Jina | `curl -s "https://r.jina.ai/网页URL"` |
+| 通用搜索引擎 | Exa (MCP) | `mcporter call 'exa.web_search_exa(query: "关键词", num_results: 5)'` |
+| 搜索小红书笔记 | 小红书 (MCP) | `mcporter call 'xiaohongshu.search_feeds(keyword: "关键词", limit: 5)'` |
+| 解析抖音视频 | 抖音 (MCP) | `mcporter call 'douyin.parse_douyin_video_info(url: "视频URL")'` |
+
+### 步骤 4：返回结果
+
+将搜索结果整理后返回给用户，包括：
+- 来源平台
+- 标题
+- 内容摘要
+- 链接
+
+## 注意事项
+
+1. 遵守平台的 API 使用规范
+2. 如需登录平台的搜索功能，确保用户已配置认证信息
+3. 如果搜索失败，尝试其他平台或告知用户可能的错误原因
+4. YouTube/Bilibili 需要提供具体的视频 URL，不能直接搜索关键词
+5. 详细的配置说明（Cookie、代理等）请参考 linkworld.md

package/dist/extensions/builtin/link-world/link-world-agent.md

@@ -1,82 +1,82 @@
----
-name: link-world
-description: Built-in internet access bridge for NanoPencil. Use when a task needs agent-reach backed web search, browsing, or remote internet actions.
----
-
-# Link-world
-
-NanoPencil exposes link-world as first-class tools. Prefer these tools over calling `agent-reach` through `bash`.
-
-## Tool Selection
-
-- Use `web_search` for ordinary search and lightweight web research, but only if the tool is available in the current session.
-- Use `web_fetch` when you already have the target URL and need its content, but only if the tool is available in the current session.
-- Use `link_world_admin` first when you need setup status, version, install help, or diagnostics.
-- Use `link_world_exec` only when you need a lower-level `agent-reach` command that `web_search` does not model.
-- If the task is about interacting with a live page rather than retrieving knowledge, switch to the `browser` tool family instead of forcing link-world.
-- Do not hand-build shell strings for `agent-reach` unless the dedicated tools are unavailable.
-
-## Recommended Flow
-
-1. Check readiness with `link_world_admin` action `status` or `doctor`.
-2. If the runtime is missing, use `link_world_admin` action `install_help` or `/link-world install`.
-3. If `web_search` is available, use it for ordinary search.
-4. If `web_fetch` is available and the exact URL is already known, use it.
-5. If high-level link-world tools are unavailable, call `link_world_exec` with explicit argv-style arguments.
-6. Prefer bundled site or domain skills before inventing a new web workflow.
-
-## Examples
-
-Admin:
-
-```json
-{
-  "action": "status"
-}
-```
-
-Execution:
-
-High-level search:
-
-```json
-{
-  "query": "OpenAI Responses API"
-}
-```
-
-```json
-{
-  "query": "Playwright file upload",
-  "provider": "web",
-  "limit": 5
-}
-```
-
-High-level fetch:
-
-```json
-{
-  "url": "https://example.com/docs"
-}
-```
-
-Lower-level execution:
-
-```json
-{
-  "args": ["doctor"]
-}
-```
-
-```json
-{
-  "args": ["search", "OpenAI Responses API"]
-}
-```
-
-## Notes
-
-- Pass `args` as an array of tokens, not as one shell command string.
-- If `agent-reach` is not installed, the tools will return install guidance.
-- Keep NanoPencil as the integration boundary: diagnostics through `link_world_admin`, search through `web_search`, direct retrieval through `web_fetch`, lower-level execution through `link_world_exec`.
+---
+name: link-world
+description: Built-in internet access bridge for NanoPencil. Use when a task needs agent-reach backed web search, browsing, or remote internet actions.
+---
+
+# Link-world
+
+NanoPencil exposes link-world as first-class tools. Prefer these tools over calling `agent-reach` through `bash`.
+
+## Tool Selection
+
+- Use `web_search` for ordinary search and lightweight web research, but only if the tool is available in the current session.
+- Use `web_fetch` when you already have the target URL and need its content, but only if the tool is available in the current session.
+- Use `link_world_admin` first when you need setup status, version, install help, or diagnostics.
+- Use `link_world_exec` only when you need a lower-level `agent-reach` command that `web_search` does not model.
+- If the task is about interacting with a live page rather than retrieving knowledge, switch to the `browser` tool family instead of forcing link-world.
+- Do not hand-build shell strings for `agent-reach` unless the dedicated tools are unavailable.
+
+## Recommended Flow
+
+1. Check readiness with `link_world_admin` action `status` or `doctor`.
+2. If the runtime is missing, use `link_world_admin` action `install_help` or `/link-world install`.
+3. If `web_search` is available, use it for ordinary search.
+4. If `web_fetch` is available and the exact URL is already known, use it.
+5. If high-level link-world tools are unavailable, call `link_world_exec` with explicit argv-style arguments.
+6. Prefer bundled site or domain skills before inventing a new web workflow.
+
+## Examples
+
+Admin:
+
+```json
+{
+  "action": "status"
+}
+```
+
+Execution:
+
+High-level search:
+
+```json
+{
+  "query": "OpenAI Responses API"
+}
+```
+
+```json
+{
+  "query": "Playwright file upload",
+  "provider": "web",
+  "limit": 5
+}
+```
+
+High-level fetch:
+
+```json
+{
+  "url": "https://example.com/docs"
+}
+```
+
+Lower-level execution:
+
+```json
+{
+  "args": ["doctor"]
+}
+```
+
+```json
+{
+  "args": ["search", "OpenAI Responses API"]
+}
+```
+
+## Notes
+
+- Pass `args` as an array of tokens, not as one shell command string.
+- If `agent-reach` is not installed, the tools will return install guidance.
+- Keep NanoPencil as the integration boundary: diagnostics through `link_world_admin`, search through `web_search`, direct retrieval through `web_fetch`, lower-level execution through `link_world_exec`.