{"skill":{"slug":"aana-message-send-guardrail-skill","displayName":"AANA Message Send Guardrail Skill","summary":"Validates message destination, audience, tone, privacy, and approval before sending or posting on Slack, Teams, Discord, SMS, or public channels.","description":"# AANA Message Send Guardrail Skill\n\nUse this skill before an OpenClaw-style agent sends, posts, replies, forwards, or schedules a Slack, Teams, Discord, SMS, DM, or public channel message.\n\nThis is an instruction-only skill. It does not install packages, run commands, write files, call services, persist memory, or send messages on its own.\n\n## Core Principle\n\nMessage sending is an external action. Verify channel, recipients, tone, private data, attachments, claims, and explicit send approval before posting.\n\n## Required Checks\n\n- destination: DM, group DM, private channel, public channel, SMS, or external workspace\n- recipients, mentions, broadcast tags, and channel visibility\n- tone, urgency, and conflict risk\n- private data, secrets, customer data, internal links, and hidden thread context\n- attachments, screenshots, links, files, and quoted content\n- factual claims, commitments, deadlines, or policy statements\n- explicit approval for sending or posting\n\n## Message Risk Classes\n\nTreat these as higher risk:\n\n- public channels, external workspaces, SMS, group DMs, broadcast tags, and large channels,\n- `@channel`, `@here`, role mentions, customer channels, executive channels, and incident channels,\n- screenshots, logs, file attachments, account details, customer details, or private thread context,\n- promises about refunds, deadlines, policy exceptions, security, incidents, legal/medical/financial matters, or roadmap commitments,\n- tense, emotional, disciplinary, urgent, or conflict-prone messages.\n\n## Destination Rules\n\nDo not send when:\n\n- the destination is inferred, ambiguous, or broader than the user requested,\n- a public channel or broadcast mention would expose private content,\n- a thread reply includes hidden context that new recipients should not see,\n- the message is intended for one person but would notify a group.\n\nAsk for confirmation when the destination or audience is not exact.\n\n## Tone And Claim Rules\n\nRevise before sending if tone is hostile, manipulative, too casual for a high-impact context, or overconfident about uncertain facts.\n\nDo not send unsupported promises, policy claims, customer facts, incident updates, deadlines, or commitments. Retrieve evidence, revise, or route to review first.\n\n## Review Payload\n\nWhen using a configured AANA checker, send only a minimal redacted review payload:\n\n- `destination_status`\n- `audience_status`\n- `tone_status`\n- `privacy_status`\n- `approval_status`\n- `message_risks`\n- `blocker_reason`\n- `safe_alternative`\n- `recommended_action`\n\nDo not include raw private messages, screenshots, logs, secrets, customer records, or unrelated thread history when a redacted summary is enough.\n\n## Decision Rule\n\n- If this is a draft only, draft and stop.\n- If channel, recipients, or approval are unclear, ask.\n- If private data or unsupported claims appear, redact or revise.\n- If sending is external, public, high-impact, or irreversible, require approval.\n- If the send is unsafe, unauthorized, deceptive, or privacy-violating, block.\n\n## Output Pattern\n\n```text\nAANA message gate:\n- Destination: dm / group / private_channel / public_channel / sms / external\n- Audience: exact / ambiguous / broad / public\n- Tone: ready / revise / conflict_risk\n- Privacy: clear / needs_redaction / sensitive / unknown\n- Approval: approved / required / unclear / denied\n- Decision: draft / revise / ask / redact / request_approval / send / block\n```\n","topics":["Message","Discord","Slack"],"tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":346,"installsAllTime":13,"installsCurrent":0,"stars":1,"versions":1},"createdAt":1777943311954,"updatedAt":1778492846193},"latestVersion":{"version":"1.0.0","createdAt":1777943311954,"changelog":"AANA Message Send Guardrail Skill 1.0.0 – Initial Release\n\n- Provides clear instructions for checking and approving messages before sending via channels like Slack, Teams, Discord, and SMS.\n- Outlines required pre-send checks, including channel, recipients, message tone, privacy, and explicit approval.\n- Defines risk classes for high-impact or sensitive messages and lists situations that require extra caution or confirmation.\n- Establishes rules to prevent sending when destination, audience, or privacy is unclear or risky.\n- Requires revision for hostile tone or unsupported claims; blocks unsafe or unauthorized sends.\n- Specifies a minimal, redacted review payload for external checkers to protect sensitive information.\n- Includes a standardized output pattern for summarizing send safety and recommendations.","license":"MIT-0"},"metadata":null,"owner":{"handle":"mindbomber","userId":"s177cynx168hg36ac5nrkb9wdn85z9dq","displayName":"mindbomber","image":"https://avatars.githubusercontent.com/u/42798111?v=4"},"moderation":null}