@blxzer/pactile 0.5.0 → 0.6.0-beta.4

package/dist/templates/common/bundled-skills/smart-search-cli/agents/openai.yaml

@@ -1,3 +1,3 @@
-display_name: Smart Search CLI
-short_description: Source-backed web research via CLI
-default_prompt: Use $smart-search-cli to research this current web question.
+display_name: Smart Search CLI
+short_description: Source-backed web research via CLI
+default_prompt: Use $smart-search-cli to research this current web question.

package/dist/templates/common/bundled-skills/smart-search-cli/examples/batch-search.md

@@ -1,98 +1,98 @@
-# Batch Search Workflow
-
-Use this workflow when the user gives multiple queries, tools, companies, documents, URLs, or comparison targets. The model should run repeated narrow CLI commands, save each output, and summarize only after reading the saved files.
-
-## Goal
-
-Keep batch work reproducible:
-
-- one evidence directory
-- one numbered output per query or URL
-- narrow commands selected by intent
-- fetched evidence before claim-level conclusions
-
-## Batch Query Discovery
-
-Use a PowerShell loop when several independent queries need the same source-discovery treatment.
-
-```powershell
-$Config = smart-search config path --format json | ConvertFrom-Json
-$EvidenceDir = Join-Path $Config.resolved_evidence_dir "YYYYMMDD-HHMM-batch"
-New-Item -ItemType Directory -Force -Path $EvidenceDir | Out-Null
-
-$Queries = @(
-  @{ Name = "openai"; Query = "OpenAI Responses API web_search documentation" },
-  @{ Name = "anthropic"; Query = "Anthropic tool use web search documentation" },
-  @{ Name = "google"; Query = "Google Gemini grounding search documentation" }
-)
-
-$Index = 1
-foreach ($Item in $Queries) {
-  $OutputPath = Join-Path $EvidenceDir ("{0:D2}-exa-{1}.json" -f $Index, $Item["Name"])
-  smart-search exa-search ($Item["Query"]) --num-results 5 --include-text --include-highlights --format json --output $OutputPath
-  $Index += 1
-}
-```
-
-After the loop:
-
-1. Read each saved JSON.
-2. Select the top URL or official result for each item.
-3. Fetch pages that will support final claims.
-4. Compare only what fetched text supports.
-
-## Batch Fetch Known URLs
-
-Use this when the user already provided URLs or when discovery produced a short URL list.
-
-```powershell
-$Config = smart-search config path --format json | ConvertFrom-Json
-$EvidenceDir = Join-Path $Config.resolved_evidence_dir "YYYYMMDD-HHMM-url-batch"
-New-Item -ItemType Directory -Force -Path $EvidenceDir | Out-Null
-
-$Urls = @(
-  "https://example.com/source-a",
-  "https://example.com/source-b",
-  "https://example.com/source-c"
-)
-
-$Index = 1
-foreach ($Url in $Urls) {
-  $OutputPath = Join-Path $EvidenceDir ("{0:D2}-fetch.md" -f $Index)
-  smart-search fetch $Url --format markdown --output $OutputPath
-  $Index += 1
-}
-```
-
-## Mixed Intent Batches
-
-Pick the command per item instead of forcing every item through one provider:
-
-- Broad/current/domestic item: run the bilingual `smart-search search` pair and save separate `*-zh.json` / `*-en.json` outputs.
-- Docs/API/library item: `smart-search context7-library "library" "topic" --format json --output PATH`, then `context7-docs` for the selected library id.
-- Official/trusted-domain item: `smart-search exa-search "query" --num-results 5 --include-text --include-highlights --format json --output PATH`
-- Broad first pass: `smart-search search "中文搜索，优先检索中文来源，并回答原问题：query" --validation balanced --extra-sources 1 --format json --output PATH`, then `smart-search search "Search English-language sources and answer the original question: query" --validation balanced --extra-sources 1 --format json --output PATH`
-- Known URL: `smart-search fetch "https://example.com/source" --format markdown --output PATH`
-
-## Summarize The Batch
-
-Before answering:
-
-1. Confirm every claim maps to a fetched file or fetched URL.
-2. Separate fetched evidence from unfetched candidates.
-3. Keep per-item gaps visible instead of smoothing them over.
-4. Include the evidence directory and important command lines when useful.
-
-Use a compact summary table:
-
-| Item | Evidence file | Supported conclusion | Gaps |
-| --- | --- | --- | --- |
-| item-a | `03-fetch-a.md` | supported claim | none |
-| item-b | `04-fetch-b.md` | partial claim | missing official source |
-
-## Guardrails
-
-- Keep `--extra-sources` small, usually `1` to `3`.
-- Do not use broad `search.content` as proof for a comparison row.
-- Do not cite `extra_sources` as verified evidence until fetched.
-- Do not silently switch to native web search if a command fails; report the failure and recovery path.
+# Batch Search Workflow
+
+Use this workflow when the user gives multiple queries, tools, companies, documents, URLs, or comparison targets. The model should run repeated narrow CLI commands, save each output, and summarize only after reading the saved files.
+
+## Goal
+
+Keep batch work reproducible:
+
+- one evidence directory
+- one numbered output per query or URL
+- narrow commands selected by intent
+- fetched evidence before claim-level conclusions
+
+## Batch Query Discovery
+
+Use a PowerShell loop when several independent queries need the same source-discovery treatment.
+
+```powershell
+$Config = smart-search config path --format json | ConvertFrom-Json
+$EvidenceDir = Join-Path $Config.resolved_evidence_dir "YYYYMMDD-HHMM-batch"
+New-Item -ItemType Directory -Force -Path $EvidenceDir | Out-Null
+
+$Queries = @(
+  @{ Name = "openai"; Query = "OpenAI Responses API web_search documentation" },
+  @{ Name = "anthropic"; Query = "Anthropic tool use web search documentation" },
+  @{ Name = "google"; Query = "Google Gemini grounding search documentation" }
+)
+
+$Index = 1
+foreach ($Item in $Queries) {
+  $OutputPath = Join-Path $EvidenceDir ("{0:D2}-exa-{1}.json" -f $Index, $Item["Name"])
+  smart-search exa-search ($Item["Query"]) --num-results 5 --include-text --include-highlights --format json --output $OutputPath
+  $Index += 1
+}
+```
+
+After the loop:
+
+1. Read each saved JSON.
+2. Select the top URL or official result for each item.
+3. Fetch pages that will support final claims.
+4. Compare only what fetched text supports.
+
+## Batch Fetch Known URLs
+
+Use this when the user already provided URLs or when discovery produced a short URL list.
+
+```powershell
+$Config = smart-search config path --format json | ConvertFrom-Json
+$EvidenceDir = Join-Path $Config.resolved_evidence_dir "YYYYMMDD-HHMM-url-batch"
+New-Item -ItemType Directory -Force -Path $EvidenceDir | Out-Null
+
+$Urls = @(
+  "https://example.com/source-a",
+  "https://example.com/source-b",
+  "https://example.com/source-c"
+)
+
+$Index = 1
+foreach ($Url in $Urls) {
+  $OutputPath = Join-Path $EvidenceDir ("{0:D2}-fetch.md" -f $Index)
+  smart-search fetch $Url --format markdown --output $OutputPath
+  $Index += 1
+}
+```
+
+## Mixed Intent Batches
+
+Pick the command per item instead of forcing every item through one provider:
+
+- Broad/current/domestic item: run the bilingual `smart-search search` pair and save separate `*-zh.json` / `*-en.json` outputs.
+- Docs/API/library item: `smart-search context7-library "library" "topic" --format json --output PATH`, then `context7-docs` for the selected library id.
+- Official/trusted-domain item: `smart-search exa-search "query" --num-results 5 --include-text --include-highlights --format json --output PATH`
+- Broad first pass: `smart-search search "中文搜索，优先检索中文来源，并回答原问题：query" --validation balanced --extra-sources 1 --format json --output PATH`, then `smart-search search "Search English-language sources and answer the original question: query" --validation balanced --extra-sources 1 --format json --output PATH`
+- Known URL: `smart-search fetch "https://example.com/source" --format markdown --output PATH`
+
+## Summarize The Batch
+
+Before answering:
+
+1. Confirm every claim maps to a fetched file or fetched URL.
+2. Separate fetched evidence from unfetched candidates.
+3. Keep per-item gaps visible instead of smoothing them over.
+4. Include the evidence directory and important command lines when useful.
+
+Use a compact summary table:
+
+| Item | Evidence file | Supported conclusion | Gaps |
+| --- | --- | --- | --- |
+| item-a | `03-fetch-a.md` | supported claim | none |
+| item-b | `04-fetch-b.md` | partial claim | missing official source |
+
+## Guardrails
+
+- Keep `--extra-sources` small, usually `1` to `3`.
+- Do not use broad `search.content` as proof for a comparison row.
+- Do not cite `extra_sources` as verified evidence until fetched.
+- Do not silently switch to native web search if a command fails; report the failure and recovery path.

package/dist/templates/common/bundled-skills/smart-search-cli/examples/evidence-gathering.md

@@ -1,89 +1,89 @@
-# Evidence Gathering Workflow
-
-Use this workflow when the user wants a source-backed answer, citation trail, or audit-friendly notes, but does not need full `smart-search research`.
-
-## Goal
-
-Answer from fetched evidence, not from broad generated `search.content` alone.
-
-## When To Use
-
-- The user asks for sources, citations, verification, or a defensible short answer.
-- The task can be answered with a small number of discovery and fetch commands.
-- Full Deep Research would be more work than the user asked for.
-
-Use `smart-search research` instead when the user asks for deep research, cross-checking, serious comparison, or a multi-stage investigation.
-
-## Evidence Directory
-
-Create one directory per user question. Keep command outputs numbered so the trail is easy to inspect.
-
-```powershell
-$Config = smart-search config path --format json | ConvertFrom-Json
-$EvidenceDir = Join-Path $Config.resolved_evidence_dir "YYYYMMDD-HHMM-topic"
-New-Item -ItemType Directory -Force -Path $EvidenceDir | Out-Null
-```
-
-## Step 1: Discover Candidate Sources
-
-Run the bilingual broad-search pair first, then add intent-specific discovery only when needed.
-
-Chinese-source broad pass:
-
-```powershell
-smart-search search "中文搜索，优先检索中文来源，并回答原问题：query" --validation balanced --extra-sources 1 --format json --output "$EvidenceDir\01-search-zh.json"
-```
-
-English-source broad pass:
-
-```powershell
-smart-search search "Search English-language sources and answer the original question: query" --validation balanced --extra-sources 1 --format json --output "$EvidenceDir\02-search-en.json"
-```
-
-Official domains, papers, product pages, or trusted sites:
-
-```powershell
-smart-search exa-search "query" --num-results 5 --include-text --include-highlights --format json --output "$EvidenceDir\01-exa.json"
-```
-
-Docs/API/library/framework intent:
-
-```powershell
-smart-search context7-library "react" "useEffect cleanup" --format json --output "$EvidenceDir\01-context7-library.json"
-smart-search context7-docs "/facebook/react" "useEffect cleanup" --format json --output "$EvidenceDir\02-context7-docs.json"
-```
-
-## Step 2: Fetch The Pages That Matter
-
-Pick the one or two URLs that directly support the answer. Fetch exact pages before making claim-level statements.
-
-```powershell
-smart-search fetch "https://example.com/source" --format markdown --output "$EvidenceDir\03-fetch-source.md"
-smart-search fetch "https://example.com/second-source" --format markdown --output "$EvidenceDir\04-fetch-second-source.md"
-```
-
-## Step 3: Write The Answer
-
-Use this evidence policy:
-
-- Cite fetched page text or fetched URLs for claims.
-- Treat `primary_sources` and `extra_sources` as discovery candidates until fetched.
-- If a useful candidate was not fetched, label it as an unfetched candidate.
-- Include the key command lines or the evidence directory path when the user may need to audit the work.
-
-## Minimal Final-Answer Shape
-
-```text
-Evidence used:
-- 03-fetch-source.md: <what it supports>
-- 04-fetch-second-source.md: <what it supports>
-
-Answer:
-<short answer grounded in fetched text>
-
-Unverified candidates:
-- <candidate URL or source from search/exa output, if relevant>
-
-Commands:
-- smart-search ...
-```
+# Evidence Gathering Workflow
+
+Use this workflow when the user wants a source-backed answer, citation trail, or audit-friendly notes, but does not need full `smart-search research`.
+
+## Goal
+
+Answer from fetched evidence, not from broad generated `search.content` alone.
+
+## When To Use
+
+- The user asks for sources, citations, verification, or a defensible short answer.
+- The task can be answered with a small number of discovery and fetch commands.
+- Full Deep Research would be more work than the user asked for.
+
+Use `smart-search research` instead when the user asks for deep research, cross-checking, serious comparison, or a multi-stage investigation.
+
+## Evidence Directory
+
+Create one directory per user question. Keep command outputs numbered so the trail is easy to inspect.
+
+```powershell
+$Config = smart-search config path --format json | ConvertFrom-Json
+$EvidenceDir = Join-Path $Config.resolved_evidence_dir "YYYYMMDD-HHMM-topic"
+New-Item -ItemType Directory -Force -Path $EvidenceDir | Out-Null
+```
+
+## Step 1: Discover Candidate Sources
+
+Run the bilingual broad-search pair first, then add intent-specific discovery only when needed.
+
+Chinese-source broad pass:
+
+```powershell
+smart-search search "中文搜索，优先检索中文来源，并回答原问题：query" --validation balanced --extra-sources 1 --format json --output "$EvidenceDir\01-search-zh.json"
+```
+
+English-source broad pass:
+
+```powershell
+smart-search search "Search English-language sources and answer the original question: query" --validation balanced --extra-sources 1 --format json --output "$EvidenceDir\02-search-en.json"
+```
+
+Official domains, papers, product pages, or trusted sites:
+
+```powershell
+smart-search exa-search "query" --num-results 5 --include-text --include-highlights --format json --output "$EvidenceDir\01-exa.json"
+```
+
+Docs/API/library/framework intent:
+
+```powershell
+smart-search context7-library "react" "useEffect cleanup" --format json --output "$EvidenceDir\01-context7-library.json"
+smart-search context7-docs "/facebook/react" "useEffect cleanup" --format json --output "$EvidenceDir\02-context7-docs.json"
+```
+
+## Step 2: Fetch The Pages That Matter
+
+Pick the one or two URLs that directly support the answer. Fetch exact pages before making claim-level statements.
+
+```powershell
+smart-search fetch "https://example.com/source" --format markdown --output "$EvidenceDir\03-fetch-source.md"
+smart-search fetch "https://example.com/second-source" --format markdown --output "$EvidenceDir\04-fetch-second-source.md"
+```
+
+## Step 3: Write The Answer
+
+Use this evidence policy:
+
+- Cite fetched page text or fetched URLs for claims.
+- Treat `primary_sources` and `extra_sources` as discovery candidates until fetched.
+- If a useful candidate was not fetched, label it as an unfetched candidate.
+- Include the key command lines or the evidence directory path when the user may need to audit the work.
+
+## Minimal Final-Answer Shape
+
+```text
+Evidence used:
+- 03-fetch-source.md: <what it supports>
+- 04-fetch-second-source.md: <what it supports>
+
+Answer:
+<short answer grounded in fetched text>
+
+Unverified candidates:
+- <candidate URL or source from search/exa output, if relevant>
+
+Commands:
+- smart-search ...
+```