Send to user
send_to_userDestructive
Communicate asynchronously with the user. Use when they say post this / make social posts / give this to my team — first clear ask, this turn; the card is the yes (do not re-ask). Create a card in their Command Center ONLY when the user's judgment changes the outcome (a real decision they must make — approve/deny/edit, connect an integration, post content, grant a capability). For FYI / progress / "I did X" use intent:'update' — it is logged to the activity feed (Team Activity), NOT a card. Do not create cards for non-decisions.
Routing: The one door for delivering work to the user; hired/scheduled agents post content ONLY via intent publish (never call post_to_x — that is the operator MCP/chat door); a card ONLY when their judgment changes the outcome, else intent update
[write-tier — first use may require a manager's approval; a from-now-on approval makes future calls seamless, a just-once approval re-asks next time. Call it on the first clear ask; the card is the yes — do not re-ask in chat.]
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| title | Yes | Scannable headline. If the user glances at it between meetings, they should know what it is and whether it matters. | |
| intent | Yes | What kind of communication is this? The bar for a CARD: create one ONLY when the user's judgment changes the outcome (a real decision). FYI / progress / "I did X" is intent:'update' — logged to the activity feed (Team Activity), NOT a card. Do not create cards for non-decisions. - decision: needs user approval/rejection. User sees Approve/Edit/Deny buttons. (CARD) - update: FYI / progress / completed work. Logged to the activity feed (Team Activity), NOT a card — the user glances at it, nothing to click. - suggestion: proactive recommendation — user can accept, skip, or discuss. (CARD) - alert: something urgent needs attention. Include what the user can do about it. (CARD) - publish: content ready to post. User sees the actual content and taps Post. MUST include content_body. This is the ONLY way to publish content — never call posting tools directly. (CARD) - blocked_on_you: you are blocked on something only the USER can do in the real world (connect an integration, grant OAuth, provide an input) — NOT "approve my work". Pair with needs_from_user; its key is the idempotency key (re-runs update the same blocker, not re-pile). type "input" renders an ANSWER BOX on the card (answer lands in company knowledge under inputs/<key> and your activity re-runs automatically) — so make the ask precise, and ask only for user-exclusive inputs. (CARD) | |
| message | Yes | Executive summary — what you did, why it matters right now, and what (if anything) the user needs to do. Put the full report in the deliverable field, not here. PLAIN-LANGUAGE FLOOR: answer "what happened + what does Approve/Deny do" in plain business English; NO raw tool names (snake_case) in the message; NEVER lead with run mechanics (iteration budgets, step counts, time limits) — mechanics go last, in one short parenthetical if they matter at all. | |
| on_deny | No | Decision cards: one sentence stating what stops (or stays the same) if the user denies (e.g. "This activity won't run again; nothing else changes"). | |
| reviews | No | Optional. Results from ICP/compliance agent reviews. Shown inline on the card so user sees review status at a glance. | |
| agent_id | No | Your agent ID. Used for activity tracking. | |
| platform | No | Required for publish intent — the CHANNEL half of the destination triple (company · account · channel). Explicit only; no silent default. X/Instagram/Facebook/Threads require that company's connection in Connections (mint fails closed if missing). Meta multi-Page needs a chosen Page. LinkedIn is attended (company page · copy & open). Instagram additionally REQUIRES media_artifact_ids (no text-only posts). | |
| priority | No | Defaults based on intent (alert=urgent, decision=high, others=normal). Override only when needed. | |
| companyId | Yes | FreedomOS company id to act within (you must be a member). Required for company-scoped tools. | |
| no_change | No | Zero-delta run flag (intent "update" only, activity runs only). Set true when this run scanned/checked and found GENUINELY NOTHING new to report — the run is then recorded to the activity ledger only: no card, no feed graduation, nothing lands on the operator (his attention is reserved for runs where something happened). The claim is verified against the run's actual tool trace: if this run performed any consequential action the flag is rejected and you must deliver normally. Never use it to hide real findings or unfinished work. | |
| media_urls | No | Optional. Array of image/video URLs to include with the post. Prefer media_artifact_ids for agent-generated media. | |
| on_approve | No | Decision cards: ONE concrete sentence stating what starts happening if the user approves (e.g. "I'll assemble a draft newsletter from your essays every Monday; nothing publishes without your release"). Rendered directly under the card's buttons — this is how the user knows what the click does. ALWAYS provide it for intent:'decision' cards from an activity run — if you omit it the system derives a generic fallback from your title, which is far less clear than a line you write. | |
| target_icp | No | Recommended for publish intent whenever the company has more than one ICP. The ICP id (from get_icps) of the audience persona this content is FOR — the customer-voice review grades AS that persona. E.g. a practitioner education post declares the practitioner ICP, not the channel-partner ICP, so it is not marked down for missing partner economics it was never meant to carry. Omit (or "auto") → the platform picks a customer-class ICP automatically; an unknown id also falls back to auto. | |
| action_link | No | Optional route path for deeper context. MUST be one of: /workspace, /dashboard, /finance, /team, /growth, /brand, /content-pipeline, /media, /leads, /smart-tools, /inbox. Do NOT invent routes. /workspace is the desk where Command Center cards land; /dashboard is the numbers home (metrics); /smart-tools is Connections. | |
| deliverable | No | The actual output of your work. A good update includes it. Attach the full report, analysis, data table, or structured output here — this is what the user reviews when they want to go deeper. It must CONTAIN the artifact itself — never a placeholder/pointer like "[see attached knowledge file]". | |
| action_label | No | Optional custom label for the primary action button on the card. Use this to write a CTA that matches what the user is actually doing — much better than the generic default. Examples: "Start Day 1-3", "View Cash Flow Report", "Review SEO Audit", "Continue with Aiko". Works with or without action_link. If omitted, the system picks a contextual default. | |
| content_body | No | Required for publish intent. The EXACT text that will be published. For social posts, this is the tweet/post text verbatim. Do NOT summarize — this IS the post. For a CAROUSEL this is JUST the caption (the slide copy lives in the attached images, not here). For platform "x": must fit X's 280-character limit as X counts it (every URL = 23, most emoji = 2) by default — OR, when this company's X connection has long-form enabled (X Premium), up to 25,000 characters (plain length). Longer posts are rejected either way. | |
| content_risk | No | Required for publish intent. "standard" = settled facts, low blast radius. "high_conflict" = diverging legal/industry interpretations exist and professionalism is the product: you MUST first fact-check via query_sme, attach the SME verdict as a reviews[] row (reviewer containing "SME"), supply claim_sources, and shape the content as status-plus-conflicting-views — flat rule-claims are rejected in this class. | |
| content_type | No | Required for publish intent. What kind of content is this? | |
| activity_name | No | Name of the activity that produced this output. Used for tracking. | |
| claim_sources | No | Strongly recommended whenever the content asserts facts (dates, rules, statistics). Up to 5 https URLs you actually grounded the claims on (regulator pages, official notices, primary sources). Shown on the approval card so the operator can verify claims are real, not hallucinated. | |
| content_stage | No | Required for publish intent. Hormozi stage vocabulary: "give" = pure free value, deliberately NO link/product/CTA (judged on trust and authority built); "ask" = a promotional touch with an offer/CTA (judged on offer clarity and likelihood to act; paced by the give:ask ratio — default 3 gives per 1 ask); "retain" = for existing customers (judged on deepened product value). Declare honestly — the customer-voice review judges the content through this lens. | |
| missing_tools | No | Optional. Exact snake_case tool names from your activity skill that you lack (or need granted) to finish the job fully. Required whenever your message flags a tool gap / incomplete KPI / thin-MVP caused by a missing tool — never impact-only. Server stamps grant path + one-line "Operator: grant …" by construction. Cap 5. | |
| reply_context | No | Strongly recommended with in_reply_to_tweet_id / quote_tweet_id. Display info about the target tweet so the approval card is self-contained: { url, author (@handle), snippet (FULL text of the target tweet, up to 560 chars — do not elide; the operator verifies the reply against it) }. | |
| scheduled_for | No | Optional. ISO timestamp for when this should be posted. If omitted, posted immediately on user approval. | |
| gate_substance | No | Optional. The FULL substance the quality gate should score AND the card shows the operator READ-ONLY — caption + slide copy + citations. Use this ONLY when the post's real substance extends beyond content_body because it lives in attached media (a carousel): set content_body to the caption and gate_substance to the tool's gate_text. It is NEVER posted and NEVER char-counted. OMIT for a plain post — the gate then scores content_body. | |
| quote_tweet_id | No | Optional, X only. Tweet id (or full status URL) this post QUOTES (retweet-with-comment — the original renders embedded under your text). Do NOT also paste the tweet URL into content_body; the embed is native. Use when the take stands alone as content; prefer in_reply_to_tweet_id for answering in-thread. | |
| routing_reason | No | Optional. One-line explanation of why you chose recipient_user_id (e.g., "routing to Alex because the KR this advances is owned by them"). Required when recipient_user_id is set so the routing decision is auditable. | |
| needs_from_user | No | Optional. If your output would be significantly better with something only the user can provide (tool connection, OAuth, infrastructure), note it here. This does NOT replace delivering your work — always deliver what you can first. type: what kind of action is needed - integration: user needs to connect a tool (PostHog, Slack, etc.) - oauth: user needs to complete an OAuth flow - infrastructure: something needs to be built (new app, new feature) - input: user needs to provide information (budget, preferences, etc.) key: normalized identifier (lowercase, underscores) — "posthog", "slack_oauth", "new_app" label: human-readable action — "Connect PostHog", "Complete Slack setup" why: one sentence on what it would unlock for this deliverable urgency: nice_to_have | significant | required blocked_by: optional — if THIS action is itself gated by another action, specify the key of the parent blocker. Example: PostHog can't be connected until the new app is built → blocked_by: "new_app" The system will chain these into a tree so the user sees the ROOT action to take first. Only tag it when YOUR deliverable measurably improves with that specific action — never echo company-level blockers from other agents' work, and never tag it on decision cards that are about approval, not integrations. | |
| email_recipients | No | Required when platform is "email" (unless a connected sender identity already names the account). One or more recipient addresses that complete the destination triple — mint fails closed without them (no "email with nowhere to send" cards). | |
| recipient_user_id | No | Optional. UUID of the user this card should be routed to. Use the Routing Context block in your system prompt to pick the right recipient (e.g., the owner of the OKR this advances, or the responsibility-domain owner for the relevant domain). If omitted, the deterministic floor will pick — typically the executive of the company. Always pair with routing_reason when set. | |
| audience_awareness | No | Optional. Schwartz awareness level of the audience this piece is written FOR (unaware → most_aware). Helps the customer-voice review judge fit for that audience instead of demanding buyer-readiness from top-of-funnel content. | |
| media_artifact_ids | No | Optional. Array of agent_artifact UUIDs (images/videos) to attach to the post. Preferred over media_urls for agent-generated content — these have storage paths, metadata, and audit trails. | |
| in_reply_to_tweet_id | No | Optional, X only. Tweet id (or full status URL) this post REPLIES to — the post lands threaded under that tweet. Use for reply-pass work: answering a question in an existing thread. Pair with reply_context so the human sees what they are approving a reply to. NOTE: X blocks APP-posted replies/quotes to any author who has not @mentioned this account (platform rule since ~Feb 2026) — such a card is born with salvage actions (post-standalone / human-posts-it) instead of Post Now. If the target post does not mention this account, consider drafting a STANDALONE take instead of a reply. |