Skills Catalog

Web Search

Quick web search — fast single-source lookup. Use for: factual questions ('what is X', 'who is Y'), current events, definitions, single-source verification, quick comparisons or lookups. Returns 5 search results + optional 0-3 page scrapes + a short synthesis with sources. Faster + cheaper than `research`. Pick this when the user wants A QUICK ANSWER, not deep research. CRITICAL: When the user asks a factual question or wants current info, you MUST call this tool. Do not answer from your training data — always verify with a search.

Apollo Prospect Search

Search and enrich B2B prospects via Apollo's API. BYO API key from the brand's enrichment vendor settings. Use when the brand has Apollo credentials and wants fast structured B2B sourcing; use browse when credentials are missing or freshness matters.

Add Brand Learning

Record a learning or insight about a brand. Learnings accumulate over time — product details, what works, audience preferences, style insights. Categories: product, audience, style, performance, feature, general. Be proactive — save learnings whenever you discover something useful. CRITICAL: When the user shares an insight about their brand, you MUST call this tool to persist it. Do not just say 'I have saved that insight' without calling the tool.

Check Job Status

Check the real-time status of a content job. Pass job_id or leave empty to check the most recent job. Returns: status (queued/processing/completed/failed), progress percentage, current step, and output path if done. NEVER guess job status — always call this.

Create Brand Kit

Create a new brand kit from provided details or extracted brand info. Stores: name, colors, fonts, tone, description, industry, target audience, key messages, website URL. A General Agent is auto-deployed for new brand kits. CRITICAL: When the user asks to create a brand kit, you MUST call this tool to persist it. Do not just say 'I have created your brand kit' without calling the tool.

Create Job

Submit a configured job to the processing pipeline. Requires a prompt describing what the content is about. The prompt should describe the TOPIC, not the config (director/duration are already set). PREFER using pipeline skills instead — they handle everything in one call. CRITICAL: When the user asks to create content, you MUST call this tool. Do not just describe what you would create without calling the tool.

Create Project

Create a project under the active brand. Projects group related work — campaigns, launches, content calendars. Goals and context flow down to assigned agents. Use when the user wants to organize work into a named initiative. CRITICAL: When the user says 'create a project', 'set up a project', or 'make a new project', you MUST call this tool. Do not just describe what the project will contain without creating it.

Deploy Agent

Deploy an autonomous AI agent that runs on autopilot. Agent types: social_manager, community_manager, sales_agent, content_writer. COLLECT ALL of these before calling: agent_type, agent_name, goal, platforms, approval_mode. Also collect archetype (strategist/creative/data_nerd/hustler) and optionally work_days, work_start_hour. When you have everything, call deploy_agent ONCE with all params. The agent will be created active immediately with a rich personality. Do NOT call deploy_agent with partial params. IDENTITY RULE: YOU are the agent being hired. Use 'I' and 'my' ('I'll post to TikTok', 'My schedule will be...'). Propose a creative name. Ask 2-3 focused questions to gather what you need, then deploy. CRITICAL: When the user is ready to deploy an agent, you MUST call this tool. Do not just say 'I have deployed your agent' without calling the tool.

Extract Brand

Extract brand identity from a website URL using AI. Analyzes the site's design, copy, and metadata to extract: name, colors, fonts, tone, description, industry, target audience. Costs 5 tokens. Use when user shares a URL and wants to create a brand kit from it. CRITICAL: When the user shares a URL and asks to extract brand info, you MUST call this tool. Do not just describe what you would extract.

Generate Performance Report

Generate a performance report for an agent — posts published, engagement, growth, insights.

Generate Script

Generate a structured video script from a prompt. The script includes slides with narration text, visual directions, timing, and capture actions. If user provides their own script, pass it as custom_script — it will be adapted to Brandstaq's format. PREFER using pipeline skills which handle script generation automatically. CRITICAL: When the user asks to write a script, you MUST call this tool. Do not just write a script in chat without calling the tool.

Get Brand Kits

List the user's brand kits with names, colors, fonts, tone, industry, target audience, website URL, key messages, and learnings. Use to find brand context before creating content. Auto-selects the default brand kit if one exists. CRITICAL: When you need brand context, you MUST call this tool. Do not guess what brands exist or what their settings are.

Get Voices

List all available voices with full details: name, gender, tone, language, description, provider model, tier, and capabilities. Grouped by model. Shows top 20 by default; pass show_all='true' for the complete list. Use when user asks 'what voices are available' or 'list voices'. CRITICAL: When the user asks about available voices, you MUST call this tool. Do not guess what voices exist.

Measure Engagement

Fetch engagement metrics for recent posts — impressions, likes, comments, shares, follower growth.

Query Agent Status

Check what a deployed agent is doing — recent activity, scheduled posts, performance metrics. Pass agent_name or leave empty for the first active agent. Respond in the agent's personality (first person). CRITICAL: When the user asks what an agent is doing, you MUST call this tool. Do not guess the agent's status.

Search Brand Knowledge

Search the brand's training documents and agent memories for specific information. Use when you need detailed context — product details, FAQs, competitor info, guidelines. Pass a query (keywords or topic) and optionally brand_kit_id. CRITICAL: When you need brand-specific information, you MUST call this tool. Do not guess or fabricate brand knowledge.

Set Voice

Set a voice for the brand or a specific agent. Pass voice_name (required). Optionally pass domain (social/content/community/default) to set per-domain, agent_name to set for a specific agent, or all_agents='true' to update all agents. Default: sets brand default voice. Use when user says 'set voice to Nova', 'use Atlas for social', 'change voice', 'use Nova for all agents'. CRITICAL: When the user says 'set voice to X' or 'use voice Y', you MUST call this tool to persist the change. Do not just say 'I have set your voice' without calling the tool.

Update Agent Goals

Manage an agent's DAILY STANDING ORDERS — durable patterns of work the agent should do every day or every week. Use for: 'reply to ~50 posts/day about automated trading', 'publish 1 thread/week on backtesting', 'DM 3 prospects/day from the prediction-markets list'. Do NOT use for time-bound campaigns with end-states (use create_project for those). Actions: add (append one standing order), remove (delete matching), edit (change ONE goal in place — pass match=a few words from the existing goal + goal=the new text), list (show current standing orders + do_rules + dont_rules), replace_all (rebuild the whole list — refuses to shrink without explicit confirmation), replace_all_confirmed (bypass the shrinkage brake — only when the user explicitly asks to overwrite everything). The aliases set and replace map to replace_all; change/reword/rephrase/modify/revise map to edit. CRITICAL: to change/reword ONE goal, use `edit` (NOT replace_all — replace_all rebuilds everything and will refuse to drop the other goals). Use replace_all only when the user wants to rebuild the ENTIRE list. When the user discusses goal changes and then approves them (e.g. 'this is fine', 'looks good', 'go with that', 'approved'), you MUST call this tool to persist the changes. Never respond as if goals were updated without actually calling the tool.

Generate Chart

Generate a complete chart (PNG) — line, bar, pie, scatter, etc. Uses matplotlib and brand colors. Best for data visualization in reports, posts, or standalone analytics. CRITICAL: When the user asks for a chart or graph, you MUST call this tool. Do not just describe what the chart would show.

Generate Pdf

Generate a complete branded PDF document — brochures, reports, proposals, invoices, whitepapers, one-pagers. ALWAYS use this tool (not create_content) when the user asks for any PDF, brochure, report, proposal, or document. Brand colors, fonts, and logo applied. Uses weasyprint to convert HTML/CSS to PDF. Produces a downloadable PDF file. CRITICAL: When the user asks for a PDF or document, you MUST call this tool. Do not just describe what the document would contain.

Generate Presentation

Generate a complete branded HTML slide presentation (self-contained, viewable in browser, exportable to PDF). Brand colors, professional layout, animation-rich design. Best for pitch decks, reports, product overviews. CRITICAL: When the user asks for a presentation or slides, you MUST call this tool. Do not just outline slides without generating them.

Generate Preview Image

Generate ONE preview image from a visual description. Single-image primitive — for a multi-image video sequence use the slideshow pipeline. Best for thumbnails, hero images, social media graphics, concept previews. Pass a detailed image_prompt describing what the image should look like. CRITICAL: When the user asks to generate an image, you MUST call this tool. Do not just describe an image without generating it.

Generate Spreadsheet

Generate a complete formatted Excel spreadsheet (.xlsx) — headers, data, formulas, charts. Uses openpyxl to create professional worksheets. Best for financial reports, data exports, and analytics summaries. CRITICAL: When the user asks for a spreadsheet, you MUST call this tool. Do not just describe the columns without generating it.

Match Voice

Find the best matching voice for a description. Say 'find a warm professional voice' or 'match a voice that sounds energetic'.

Preview Voice

Generate an audio preview of a voice. Say 'preview Nova' or 'what does Atlas sound like'.

Set Agent Schedule

Change when the agent wakes up and runs. Accepts wake_time (hour 0-23), days (list or 'daily'), timezone.

Set Approval Mode

Control how the agent handles content approval: auto_publish, review_first, or notify_only.

Set Token Budget

Set the agent's daily or monthly token budget limit.

Voice Configuration

Show current voice settings for the brand and its agents. Say 'what voice am I using' or 'voice config' or 'voice status'.

Facebook

Facebook Page operations: publish text/photo/video posts, fetch metrics, get profile.

Instagram

Instagram (Graph API) operations: publish image/video posts and carousels, fetch metrics, mentions, comments, get profile, reply.

LinkedIn

LinkedIn API-based operations: publish posts/articles/carousels, fetch metrics, fetch comments, get profile, and reply via API. For browser-based actions (browsing the feed, engaging in the browser, scraping the UI) use the `browse` skill instead.

Reddit

Reddit operations: publish self/text/media submissions, reply to comments, fetch metrics, get profile.

TikTok

TikTok API-based operations: publish video, fetch metrics, and get profile. For browser-based actions (browsing, engaging in the browser, scraping the UI) use the `browse` skill instead.

Twitter / X

Twitter (X) API-based operations: publish tweets/threads/carousels, fetch metrics, fetch mentions, get profile, and reply via API. For browser-based actions (browsing the timeline, engaging in the browser, scraping the UI) use the `browse` skill instead.

YouTube

YouTube operations: publish video, fetch metrics, get channel profile.