AzureAgentForge

Query the Honcho persistent memory store for facts about peers (users or agents). Use this for any "what do you know about me / the user" style question instead of fabricating from training data.

Delegate tasks to other agents in the team by creating real PaperClip child issues. Covers single-agent delegation and multi-agent coordination, with failure handling and idempotency. Trigger phrases: any task outside your direct capabilities (email/calendar/drive) that should be routed to Coder, Infrastructure, Strategy, Planner, Researcher, Curator, QA, Security, CostGuardian, Coach, Business, or Psychology — or any task that says "coordinate", "delegate", "split this", "create child issues", or "assign to".

Semantic / neural web search via the Exa API. Use for conceptual, exploratory, or "find things like X" research where keyword search underperforms — e.g. "papers about agent memory", "companies similar to <X>", "blog posts arguing <position>". Trigger phrases: "search for", "find research/papers/articles on", "what's out there about", "find similar", "semantic search". Complements brave-search (use brave for fresh/news/exact-term). The `exa-search` CLI is on `$PATH` and emits the same JSON shape as brave-search.

Pull a plain-text transcript from a video URL (YouTube, Vimeo, Bilibili, and ~1800 other sites) so you can read, summarize, quote, or research its content without watching it. Use whenever a source or request is a video link. Trigger phrases: "summarize this video", "what does this talk/video say", "transcript of <url>", "get the captions", or any task pointing at a youtube.com / youtu.be / vimeo / video URL. The `video-transcript` CLI is on `$PATH`.

Read a web page as clean Markdown instead of raw HTML. Use whenever you need the actual content of a specific URL — an article, vendor/pricing page, changelog, docs page, or any link a search result pointed you to. Trigger phrases: "read this page", "what does <url> say", "fetch/open this link", "get the article", "extract the content", or any time you would otherwise `curl` a URL to read it. The `web-read` CLI is on `$PATH`.

How to do web research with the right tool and a fallback when one fails: which search backend to use, how to read a page cleanly, how to get a video's transcript, and how to check that the backends are healthy. Use at the start of any research/lookup/"find out about X" task, or when a research tool returns empty/erroring results and you need the fallback. Pulls together exa-search, brave-search, web-read, video-transcript, and the research-doctor health check.