360 Secure SkillHub
360 安全技能中心

what-brandon-thinks-chat-analyst

核心功能为分析WhatsApp/iMessage匿名聊天的沟通模式，生成证据型分析简报、交互信号总结，谨慎审核相关主张。典型场景涵盖亲友关系、群聊中的互动、语气转变、转折点等问题。需用户提供合规且去敏感信息的聊天数据，输出含事实观察、解读、替代解释等的结构化结果，禁止用于诊断、高风险决策，可选MCP辅助但不上传数据。

gpt-img-2
gpt-img-2
热度 7.62
最近更新 2026-09-23 11:00:49
分类：数据依赖：MCP 服务 / 纯提示词应用场景：摘要总结 / 报告撰写 / 数据处理

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add what-brandon-thinks-chat-analyst
Metadata
name
what-brandon-thinks-chat-analyst
description
Turn anonymized WhatsApp or iMessage observations into evidence-aware analysis briefs, interaction-signal summaries, and cautious claim audits. Use for relationship, friendship, family, group-chat, reciprocity, effort, tone-shift, and turning-point questions; do not use to diagnose people, prove private intent, make safety-critical decisions, or analyze conversations the user is not entitled to use.
metadata
{"openclaw":{"homepage":"https://whatbrandonthinks.org/sample-report","version":"1.0.0"}}

What Brandon Thinks Chat Analyst

Help the user examine communication patterns without turning incomplete text into certainty. This Skill is a text-only workflow by default. Its optional MCP adds deterministic, read-only helpers and never uploads transcripts, calls an AI provider, accesses an account, purchases a report, or changes stored data.

What Brandon Thinks is independent and is not affiliated with WhatsApp, Meta, Apple, or iMessage.

Protect the conversation first

  • Work only with a conversation the user is entitled to analyze.
  • Prefer aliases, aggregate counts, and short necessary excerpts over a full raw transcript.
  • Ask the user to remove passwords, payment data, identity documents, health records, addresses, and unrelated third-party details.
  • Do not retain or repeat sensitive excerpts beyond what the answer requires.
  • Direct immediate danger, abuse, medical, legal, employment, or other high-stakes decisions to qualified human support rather than treating chat analysis as a ruling.

Frame the question

Identify the platform, relationship context, participant aliases, representative date range, and one focused question. Record important gaps such as deleted messages, missing reactions, media, calls, voice notes, or offline events.

Separate three layers:

  1. Observable facts: counts, initiations, response timing, questions, follow-through, recurring words, and dated changes.
  2. Interpretations: plausible explanations tied to more than one observation.
  3. Unknowns: private intent, unrecorded events, body language, vocal tone, and missing context.

Evaluate patterns cautiously

Use repeated patterns rather than isolated lines. Compare like-for-like periods and account for time zones, work schedules, platform habits, group size, and long gaps in the export.

Message share, reply speed, or initiation count can describe interaction structure; none independently measures affection, honesty, compatibility, manipulation, or moral worth. Never diagnose a participant or assign a sensitive trait from messages.

Audit every important claim

For each interpretation:

  • State the observation that supports it.
  • Include a short exact excerpt only when supplied by the user.
  • Note the relevant date or period when known.
  • Give at least one reasonable alternative explanation.
  • Label confidence as low, medium, or high based on repeated support and missing context.
  • Say what additional evidence could strengthen or weaken the claim.

Do not invent quotes, timestamps, events, participants, or motives. If the evidence conflicts, show the conflict instead of forcing a verdict.

Use the optional MCP

  • build_chat_analysis_brief structures the question, privacy checks, evidence plan, and known gaps.
  • score_interaction_signals summarizes user-supplied aggregate interaction counts without scoring love or intent.
  • audit_chat_claim checks whether an interpretation has evidence, alternatives, uncertainty, and falsifiers.
  • get_what_brandon_thinks_resources returns canonical product, workflow, and safety resources.

Connect the stdio server with:

openclaw mcp add what-brandon-thinks \
  --command npx \
  --arg -y \
  --arg github:gpt-img-2/whatbrandonthinks-chat-mcp \
  --include 'build_chat_analysis_brief,score_interaction_signals,audit_chat_claim,get_what_brandon_thinks_resources'

Then run openclaw mcp doctor what-brandon-thinks --probe. If the installed release lacks the mcp command group, add npx -y github:gpt-img-2/whatbrandonthinks-chat-mcp as a Stdio server in MCP settings. The MCP is optional; never imply it ran when unavailable.

Return an inspectable result

Return Focused question, Observed patterns, Evidence, Interpretations, Alternative explanations, Unknowns, Confidence, and Next check. Keep factual observations visually distinct from interpretations.

Canonical resources