meeting-intelligence

Plugin meeting-intelligence from marketplace joystream-skills in joystream-ai/joystream-skills. Commands read from the repo's manifest — 2 days ago.

Skills (44)

• account-discovery-candidates

  Find new companies that plausibly fit an ICP in the directory pages and search results an agent gathered (startup directories, job boards, event attendee or sponsor pages, web, news, Hacker News, Reddit), and return a de-duplicated list of candidates, each with a one-line cited fit reason and its so

• account-icp-definition

  Turn a founder's loose sense of their target customer into a structured, versioned ICP document through an interview, one question at a time: first reads the company's own website and skips what it already answers, then asks only for what is missing (target pain, buying triggers, budget authority, r

• account-icp-qualification

  Judge how well one company fits an ideal customer profile (ICP) from the web evidence gathered about it, and return a cited fit score (0-5) with a Strong / Possible / Weak tier, a separate warmth read, a Pursue / Research / Reject verdict, a likely problem, the missing information and the shared ite

• account-icp-revision

  Propose a revision to an ideal customer profile (ICP) from real outcomes, for a person to approve: reads the versioned ICP store (the current version is the newest active row), the accept and reject decisions people recorded on prospected accounts, customer insight rows and optional call learnings,

• account-icp-rubric

  Turn an ideal customer profile (ICP) into one fixed scoring rubric, before any company is judged, so every company in a list is scored on the same criteria, kinds and weights. Returns the rubric in the shape account-icp-qualification takes as its rubric input. Use when a fan-out agent qualifies a li

• account-lookalike-profile

  Describe a good customer (the seed) along four dimensions, fit, behavior, pain and timing, every trait cited from the evidence gathered about it and from the seller's own reason it is a good customer, then turn the profile into lookalike searches led by behavior and pain (not just industry and size)

• account-research-brief

  Write a short, cited pre-call brief on one company from the web evidence gathered about it: what they do, dated recent events inside a lookback window, the likely owner of the problem you solve (always flagged as a guess), angles for outreach and open questions, plus a paste-ready brief of at most 1

• account-trigger-events

  Find dated, sourced trigger events for one company in a lookback window (funding, hiring, launches, leadership changes, partnerships) from news, search and careers-page evidence, and say why each matters. Drops undated, out-of-window and generic items and counts them. An optional rubric of the signa

• channel-message-render

  Render a channel-neutral report (the `report` every JoyStream task skill returns) as the message content for one channel: Slack text and Block Kit blocks, a Microsoft Teams HTML message, a Discord message with embeds, an HTML email draft, or plain markdown, plus a plain-text fallback. Applies each c

• community-buyer-signals

  Find people on Hacker News (and, when reachable, Reddit) who are asking for or struggling without a product like yours: a sandbox script searches the free, keyless Hacker News search API for your queries, and the model scores each post for buying intent, a present pain and vendor promotion (a probab

• company-competitor-changes

  Report what one competitor changed in the last N days (pricing, launches, positioning, funding, leadership, partnerships, hiring, customers) from news and search results and its extracted changelog, pricing and blog pages. Every change is dated, sourced and has a "so what" line against your own posi

• company-website-profile

  Profile a company from its own website and a web search: positioning, target customer, products, pricing, proof points and competitors, every field cited or marked unknown. Competitors come only from search results, never from model knowledge. Given the previous run's saved state, it also says what

• content-ai-slop-annotate

  Check one piece of writing (a social post, a blog draft, an email) for AI-writing tells and return the same text annotated in place as Markdown: each tell struck through where it occurs, with a concrete rewrite and its category from a 13-category taxonomy (false contrast, rule-of-three, formulaic tr

• content-ai-slop-check

  Score one or more pieces of written content (a social post, a blog draft, an email) for AI slop on a 1-10 scale, where 10 reads as written by a specific person and 1 reads as generic machine text: quotes each offending span, names the pattern (hype words, stock openers, rule-of-three lists, hedging,

• content-editorial-plan

  Plan next week's content calendar (date, platform, topic, angle, format and a reason for each slot) from what other agents already produced: a content performance report (what worked), market intelligence digests (what competitors and the market just did) and customer insights (pains in customers' o

• content-file-save

  Save a document an earlier step wrote (a Markdown report, a CSV table, JSON data, plain text or an HTML page) as a downloadable file in the run's output folder, exactly as written, with a safe file name, a format check and its size and checksum. Use when an agent must "make the report downloadable",

• content-news-angles

  Newsjacking: from fresh news search results (one search per topic), optional market-intelligence channel messages and the list of stories already covered, pick the few timely stories where the author has a real point of view and suggest an angle for each: why it is timely, the author's belief it con

• content-performance-review

  Weekly review of what worked in an author's social posts and why, over time, from a metrics store with one row per post per capture (impressions, reactions, comments, reposts, clicks): this period's posts against a baseline of earlier ones, the top and bottom posts with the likely reasons grounded i

• content-publish-receipt

  Write the receipt after an agent publishes approved social posts (LinkedIn, X): for each platform it says whether the post went out, the post id and its url built only from the id the platform returned, and whether the platform's echo of the posted text matches the approved text, plus spreadsheet ro

• content-social-post-draft

  Turn one piece of source content (a blog post, changelog, release notes, a page's text, or a post-worthiness verdict with story angles) into draft social posts for LinkedIn and X, one distinct angle per platform, in the author's voice: from a voice profile (made once with content-voice-profile, or r

• content-voice-profile

  Build a reusable written voice profile for one author from their own writing (posts, blogs, emails, pasted or read by an earlier step), the authors whose style they like, and what they never want to sound like: person and tone, sentence rhythm, words they use and avoid, how they open and close, emoj

• customer-insight-extract

  Turn one customer call transcript into anonymized insights (pain, theme, verbatim quote, speaker role, segment) ready to append to a Notion database or Google Sheet, and count how many calls share each pain using the rows already in that store, so repeated pains are flagged (default: 3 or more calls

• github-activity-counts

  Count the pull requests merged and opened per repository from GitHub search results and write the totals as one exact sentence. Runs a deterministic Python script in the sandbox (the agent needs code_execution), so the arithmetic is never left to the model. Use before github-repo-activity-summary or

• github-activity-post-worthiness

  Decide whether a repository's recent GitHub activity (merged pull requests, releases, closed issues in a time window) is worth a social post, and if so which story angles to use, each backed by the activity it rests on. "Nothing worth posting today" is a normal answer. Use when an agent turns GitHub

• github-project-ticket-summary

  Summarize ticket movement on a GitHub Project over a lookback window — what closed, what is in progress, what got opened — with a one-line gist of the actual work behind each closed ticket instead of just its title. Returns a markdown digest and the shared channel-neutral report for channel-message-

• github-release-digest

  Summarize a GitHub repository's releases published between a start and end date as a customer-friendly release announcement, written as one colored Block Kit Slack message for the agent's post step. Use when an agent needs a release roundup or release notes digest for a repo over a date range.

• github-release-notes-draft

  Draft customer-facing release notes for one GitHub release from the bodies of the pull requests merged since the previous tag, grouped into Action needed, New, Improved, Fixed and Security, with internal-only changes skipped and counted. Use when an agent needs release notes, a changelog entry or a

• github-repo-activity-summary

  Summarize what changed in one or more GitHub repositories over a lookback window — merged PRs, opened PRs, and notable commits — as a short, human-readable digest grouped by theme rather than a raw dump of titles. Returns a markdown digest and the shared channel-neutral report for channel-message-re

• list-prepare

  Prepare a list of items (companies, people, meetings, any kind), given as objects or as pasted text or a CSV file's text, for a fan-out agent: normalize each item's identity key, drop duplicates and excluded items, cap the list, write one search query per item, and lay the accepted items out in a fi

• meeting-huddle-summary

  Summarize one internal team huddle, stand-up or sync from its transcript (Zoom VTT, Google Meet transcript text, or "Speaker: text" lines): TL;DR bullets first, then decisions, action items with owners taken only from the attendees (or "unassigned") and due dates only when stated, blockers, per-pers

• meeting-prep-brief

  Write a short, cited brief for one upcoming external meeting from the meeting item and the evidence earlier steps gathered: a TL;DR first, then who is coming (role, with a same-name identity check), what the company does, dated recent events, history with us (only from email, CRM or past-meeting rec

• meeting-prep-select

  Pick the upcoming external meetings to prepare for from calendar events an earlier step read (Google Calendar or Outlook shapes): keep meetings that start inside a time window after now and have attendees outside your own domains, drop internal-only, declined, cancelled, all-day and non-meeting even

• outreach-message-draft

  Draft one first-touch outreach message to one person, a cold email (at most 120 words) or a LinkedIn DM (at most 300 characters), built only on the evidence you paste or an earlier research step produced (an account brief, trigger events), in the voice of your own sample messages. Every hook traces

• person-buyer-qualification

  Decide whether one person is worth reaching about the problem you solve: do they own that problem in their current role, are they publicly active (dated posts, talks, podcasts, articles, a job change) in a recent window, and are they reachable in public places where they engage. Scores a weighted ru

• person-public-profile

  Build a professional public profile of one named person from web search results: which sources are really about them (identity check on company or role), the topics they speak or write about, and views they have stated in public, each cited. Rejects same-name strangers and counts them; never returns

• person-role-finder

  Find who holds given roles at one company (e.g. VP Marketing, Head of Demand Generation, CFO) from team pages, press releases, conference speaker lists, podcasts and optional enrichment records. Each person comes with a title, the target role it fills, a dated source, a confidence and a freshness fl

• person-warm-paths

  Find warm introduction paths to target companies and people from the user's own network data: a professional-network connections export the user imported into a spreadsheet, their past calendar meetings, and optionally their email contacts. For each target it lists who the user already knows there o

• platform-alert-diagnosis

  Diagnose one JoyStream platform alert from js-monitor. Reads the alert event, makes a small, bounded set of read-only Railway calls, and posts ONE diagnosis (cause, redacted evidence, one safe recommendation, confidence) as a single reply in the alert's Slack thread. Use when the JoyStream ops agent

• product-weekly-intelligence

  Write a weekly product intelligence report that links three things: the pains customers keep naming (counted from an anonymized customer insight store), what competitors changed this week (from the channel where a competitor-watch agent posts its digest), and what engineering shipped (merged pull re

• results-rank

  Combine the per-item results of a fan-out agent into one ranked list: read every branch's item_result, order the items by the agent's rank_by rule, keep the top N, renumber the branches' sources into one list, and name every failed branch by its item. Runs a deterministic Python script in the sandbo

• sales-call-debrief

  Debrief one sales or discovery call from its transcript: key points first, then a short summary, the customer's pains and objections, verbatim quotes, next steps with owners taken only from the attendees (or "unassigned"), what was said about each topic you ask it to listen for (budget, competitors,

• skill-conformance-fix

  Draft the judgment fixes for a non-conformant Agent Skill from its files and its validator findings: a vague or missing description, a body that tells the model to fetch, a missing or broken output schema written from the body's Output section, unparseable frontmatter, or a nonstandard key the mecha

• team-weekly-brief

  Turn a week of posts from several team or agent channels (pipeline digests, content reports, product and market updates, people's own posts) into one brief for a leader: the highlights first, then what needs a decision, what is at risk, and a short section per area, every line cited to the message i

• website-aeo-geo-audit

  Score how ready a website is to be found, understood and cited by AI answer engines (ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews) — an AEO/GEO audit across six categories with strengths, gaps, prioritized fixes, a full markdown report and a short summary ready to post. Optionally checks