research

Plugin research from marketplace plugin-research-powerpack in yigitkonur/plugin-research-powerpack. Commands read from the repo's manifest — 18 days ago.

Skills (5)

• ask

  Answer one research question from web sources with verbatim, cited evidence using the Research Powerpack. Use when the answer needs the web, for example "look up whether X supports Y", "what do the docs say about", "read these URLs and tell me", what changed in a release, or an error message to diag

• deep

  Run deep, multi-source web research that ends in files on disk: a researched report, decision memo, literature review, migration guide built from web sources, or a corpus comparing five or more options. Use only when the user explicitly asks for deep research, a researched report, a landscape or mar

• doctor

  Diagnose the research plugin. Checks that the Research Powerpack server is reachable, its tool contract matches the plugin, the settings in effect, and whether the same server is configured twice. Use when research tools are missing, fail, or behave oddly.

• method

  Evidence discipline for researching on the web with the Research Powerpack MCP (plan-research, web-search, extract-evidence). Loaded by the research skills and agents; do not invoke it directly. It covers how to query, which results count, how to retry, and how to cite verbatim quotes. For a user's

• verify

  Fact-check text the user supplies, such as a file, README, docs page, PR description, blog draft, pitch or AI-written answer. Extracts the checkable claims and tests each against verbatim quotes, marking it supported, contradicted, outdated or unverified. Use when the user hands over a file, URL or

Subagents (3)

• analyst

  Use this agent to synthesize research that has already been gathered: draft one part of a report, or build one comparison file for a corpus, from existing notes and the evidence ledger. It has no web access and cannot ask the user questions. <example> Context: A deep research run has finished its sc

• auditor

  Use this agent to check whether specific claims are actually supported by the verbatim quotes they cite, or to test claims from an outside document against web sources. It judges each claim and writes a verdict table. It cannot ask the user questions. <example> Context: A deep research draft is comp

• scout

  Use this agent to research one bounded question, lens or entity on the web with verbatim, cited evidence, and report back without filling the main conversation with search output. It cannot ask the user questions; it returns open decisions instead. <example> Context: The ask skill runs with --backgr

MCP servers (1)

• powerpack

  connects to: ${RESEARCH_MCP_URL:-https://research-mcp.yigitkonur.com/mcp}

Hooks (6)

• PreToolUse

  node "${CLAUDE_PLUGIN_ROOT}/scripts/hook.mjs" guard

• PreToolUse

  node "${CLAUDE_PLUGIN_ROOT}/scripts/hook.mjs" writeguard

• PostToolUse

  node "${CLAUDE_PLUGIN_ROOT}/scripts/hook.mjs" record

• PostToolUse

  node "${CLAUDE_PLUGIN_ROOT}/scripts/hook.mjs" cite

• SessionStart

  node "${CLAUDE_PLUGIN_ROOT}/scripts/hook.mjs" session

• Stop

  node "${CLAUDE_PLUGIN_ROOT}/scripts/hook.mjs" stop

Similar plugins