benjaminjackson

Generate answers to questions with structured output using AI search and synthesis. Use when you need factual answers with citations from web sources, or when you want to extract specific structured information in response to a query.

Get code context from repositories with examples and documentation. Use when you need code snippets, implementation examples, API usage patterns, or technical documentation for programming concepts, frameworks, or libraries.

Find web content similar to a given URL using AI-powered similarity matching. Use when you have an example page and want to discover related articles, papers, or websites with similar content, style, or topic.

Retrieve and extract content from URLs with AI-powered summarization and structured data extraction. Use for scraping web pages, extracting specific information, summarizing articles, or crawling websites with subpages.

Search the web for content matching a query with AI-powered semantic search. Use for finding relevant web pages, research papers, news articles, code repositories, or any web content by meaning rather than just keywords.

Use when the user mentions Exa research OR when the workflow benefits from complex, multi-step research and other exa-ai approaches are not yielding satisfactory results.

Use when setting up monitors - periodic searches to add new items or refresh existing items in a webset automatically.

Use for creating websets, running searches, importing CSV data, managing items, and adding enrichments to extract structured data.

Use this skill whenever you are writing or updating technical documentation: a README, an install guide, setup or config docs, dev docs, API docs, help text, a troubleshooting section, and also every commit message and pull request description. Trigger on "write a README", "document this", "update the install guide", "our docs are out of date", "write the commit message", "open a PR", or "write the PR description". It edits the file directly, and it verifies every path, command, and flag against the actual code before writing about them. This is technical writing: transactional prose, no author's voice to preserve. Do NOT use it for writing that has a voice: essays, blog posts, emails, marketing copy, anything persuasive or personal. Those belong to the `draft` plugin. Do not use it to ghostwrite something new from nothing; it documents systems that already exist.

Use this skill whenever someone wants feedback on a piece of writing without committing to changing it: essays, emails, board policies, internal docs, marketing copy, blog posts, or any prose draft. Trigger on phrases like "critique this", "how does this read?", "what's wrong with this", "give me notes on this", "review this draft", "does this land", "get me a second opinion on this", or "/critique [file]". This is a read-only pass — it never edits the file. If the user wants the writing actually fixed, not just assessed, use `tighten` instead.

Use this skill whenever someone hands you a piece of writing and wants it actually fixed: tightened, cut for economy, given a rigorous line edit. Essays, emails, board policies, internal docs, marketing copy, blog posts, or any prose draft. Applies every edit automatically, no approval step (at most a one-line calibration question on a long doc whose purpose is ambiguous) — the right call whenever the ask is "just make this better" and the user doesn't want to review each change. Trigger on phrases like "tighten this", "clean this up", "fix this", "edit this for economy", "make this leaner", "cut the fluff from this doc", "give this a line edit and apply it", "this feels bloated, fix it", or "/tighten [file]". Works on a single document or a batch in one run: several board policies, a folder of drafts, a stack of emails. If the user wants feedback without the file being changed, use `critique` instead. Don't reach for this when the request is to write something new, brainstorm, or restructure unprompted; it e

Simulates a page-one editorial meeting on a pitch or idea: a dynamic panel of editor personas debates it live via SendMessage, then delivers a verdict. Trigger on "page one meeting", "run this by the editors", "simulate a pitch meeting", "editorial meeting on this", "get the room's take", "workshop this with a panel of editors", or /page-one-meeting. For a single quick editorial read instead of a multi-agent debate, use the managing-editor agent directly, not this skill.

Runs a pitch through a single managing-editor read instead of the full page-one-meeting panel: one Socratic verdict (green light / revise / kill), no cast, no debate, no dissent round. Trigger on "sharpen this pitch", "sharpen my pitch", "quick editorial read", "give me an editor's take", "run this by an editor", or /sharpen. For a full multi-persona debate instead, use the page-one-meeting skill.

Use when facing a complex, multi-part problem that would benefit from being broken into independent and dependent sub-questions before answering, rather than reasoned about as one tangled whole. Trigger on "atom of thought", "AoT", any request to "break down" or "break this down" (a problem, question, or decision), any request to reason or think "step by step", "decompose this problem", "atomic decomposition", or any complex analytical, planning, or reasoning task where a direct one-shot answer risks missing a sub-part or an implicit dependency between parts.

Run an engineering postmortem on a just-finished task or session — what was learned, where time was wasted, and the concrete fixes so it goes faster next time. Reconstructs the facts, hands them to ONE well-briefed reviewer subagent, and returns a root-cause analysis, a minimum-round-trip replay, and proposed edits to the skill/config/code that caused the friction. Use when the user says "postmortem", "retro", "what did we learn", "how could that have gone better", "how do we do this more efficiently next time", or asks to review how a piece of work went.

Use when you've just completed work and the user wants an honest audit of it — confidence levels, shortcuts taken, unverified assumptions, and known weaknesses. Triggers on: "confess", "be honest about this", "what are you unsure about", "what did you cut corners on", "what's the weakest part", "audit your own work", "what assumptions did you make", or after any significant output where the user wants unvarnished reflection rather than reassurance.

Use when the user is writing, planning, or reviewing a recurring email newsletter for a nonprofit aimed at cultivating and retaining donors. Trigger on: "donor newsletter", "draft a newsletter", "newsletter cadence", "newsletter subject line", "newsletter layout", "email deliverability", "open rate", "newsletter A/B test". Focuses on the newsletter as a cultivation/stewardship instrument, distinct from one-off appeals and broader stewardship strategy.

Use when writing a donor acknowledgment letter or gift receipt, planning a stewardship communication calendar, building a donor impact report, running moves management for a major donor, stewarding a planned/legacy gift, reporting to a corporate or foundation funder, re-engaging a lapsed donor, running digital/multichannel stewardship, tracking retention or stewardship ROI metrics, or stewarding a grateful patient/constituent program. Trigger on: "how do I steward", "donor retention", "acknowledgment letter", "moves management", "lapsed donors", "impact report", "planned giving", "grant reporting".

This skill should be loaded when working with Ruby files to understand when to use LSP operations (documentSymbol, findReferences, goToDefinition, hover) versus standard tools (Read, Grep). Critical for preventing unnecessary file reads when structural information is needed.

Manage tasks in Reclaim.ai calendar scheduling app. Use when creating, updating, listing, completing, or deleting Reclaim tasks, or working with calendar scheduling, task priorities, time blocking, or task duration management.

Ask the Oracle — a one-shot Fable answer from a stripped headless session (~1.3k tokens of overhead vs ~6k for the oracle subagent). Use when the user says "ask the oracle", "oracle,", or "/oracle <question>".

Turn an article URL into a polished, editorial share card: the article's own hero image, a pulled quote as the dominant element, plus title, byline, outlet name and favicon. Use when the user pastes an article link and wants a share graphic, quote card, or social card. Invoke explicitly as `/liftout:create <url> [aspect-ratio]`, where aspect-ratio is portrait (default), square, or landscape. Trigger on "share card", "quote card", "liftout", "make a card from this article", "social graphic for this link".

Use this skill when creating, optimizing, or improving prompts for large language models. Applies when users need help designing effective prompts, selecting appropriate prompting techniques, or troubleshooting prompt performance. Provides expert guidance through conversational consulting to build prompts using research-backed best practices and proven patterns.