Stats
Actions
Tags
From greedysearch
Multi-engine web search and deep research with Perplexity, Google, Gemini, ChatGPT, Semantic Scholar, and Logically. Supports all-engine grounded search, optional synthesis, and iterative research workflows.
How this skill is triggered — by the user, by Claude, or both
Slash command
/greedysearch:greedy-searchThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
`greedy_search({ query, engine: "all"|"perplexity"|"google"|"chatgpt"|"gemini"|"semantic-scholar"|"logically"|"bing", synthesize?: bool, synthesizer?: "gemini"|"chatgpt", depth?: "research", breadth: 1-5, iterations: 1-3, maxSources: 3-12, researchOutDir?: string, writeResearchBundle?: bool, visible: bool })`
greedy_search({ query, engine: "all"|"perplexity"|"google"|"chatgpt"|"gemini"|"semantic-scholar"|"logically"|"bing", synthesize?: bool, synthesizer?: "gemini"|"chatgpt", depth?: "research", breadth: 1-5, iterations: 1-3, maxSources: 3-12, researchOutDir?: string, writeResearchBundle?: bool, visible: bool })
Modes: individual engine search · grounded engine:"all" search with fetched sources · optional synthesize:true using the configured synthesizer over all-engine results · depth:"research" for the iterative deep-research workflow.
Config: ~/.pi/greedyconfig supports { "engines": ["perplexity", "google", "chatgpt", "gemini"], "synthesizer": "gemini" } by default. semantic-scholar and logically are opt-in academic/research engines — add them to engines only when you want academic paper discovery or research-assistant workflows in the normal all-search fan-out. Without explicit opt-in, engine:"all" excludes them because their results are noisy for casual web search; they shine in depth:"research" mode. Any configured engine can participate in engine:"all"; deep research child searches reuse the same configured engines list and stdin-safe query passing. Normal all-search synthesis remains controlled separately by synthesizer; research planning/final synthesis uses Gemini.
Compatibility: legacy depth:"fast"|"standard"|"deep" is still accepted. fast skips source fetching; standard/deep alias synthesize:true. Prefer synthesize:true, optional synthesizer, and depth:"research" going forward.
Research output: depth:"research" writes a dataroom-style bundle by default under .pi/greedysearch-research/<timestamp>_<query>/ with STATUS.md, OUTLINE.md, reports/SUMMARY.md, reports/CLAIMS.md, reports/GAPS.md, sources/, and data/manifest.json. Pass researchOutDir to choose the directory or writeResearchBundle:false to disable disk output.
Scale-aware research: When breadth and iterations are not explicitly set, the classifier auto-detects query complexity. Simple queries ("what is X") use a fast single-pass path (~70% faster). Moderate queries get tighter breadth/iterations. Complex queries use the full loop. Explicit breadth/iterations always override the classifier.
Auto-recovery: Headless default. Bing/Perplexity auto-retry visible on CF block. Manual CAPTCHA → visible stays open; solve then rerun.
CDP safety: Use bin/cdp-greedy.mjs only. Never raw bin/cdp.mjs.
npx claudepluginhub apmantza/greedysearch-pi --plugin greedysearchOrchestrates parallel web search, browser-explorer, and Gemini CLI agents to produce cited research reports. Invoke with /deep-research <topic>.
Conducts deep web research with parallel agents, multi-wave exploration for gaps, and structured synthesis. Activates for investigating topics, comparing options, best practices, or comprehensive web info.
Researches complex topics across multiple sources with configurable depth, focus, and output format via interactive questions.