This article covers five new Promptolis originals that share a common thread: where most AI-prompt content fails on HONESTY, and the Truth Series wedge is sharpest. Salary negotiation guides give you motivational fluff. Speech-writing tools produce generic templates. Customer-service scripts are saccharine corporate-speak. MCP coverage is hype-drunk or skeptic-cynical. AI agent design is "be autonomous!" without specifications.
The Truth Series approach: real leverage assessment, specific behavioral output, evidence-based frameworks, refusal of platitudes. We'll cover what each of the five new originals does + when each fails + how they fit together.
The Five New Truth Series Originals
- Salary Negotiation: Honest Scripts (No Apology, Real Leverage)
- Speech Writer: Wedding, Eulogy, Toast, Retirement
- Customer Service De-Escalation Scripts
- MCP Server Prompt Engineering
- AI Agent Workflow Architect
These five share a pattern: high search volume, weak honest competition, perfect Truth Series brand-fit.
Salary Negotiation: The 60-Second Silence Rule
Most salary-negotiation guides fail with motivational platitudes: "Believe in your worth!" "Don't sell yourself short!" "Confidence is key!" None of these tell you what to actually SAY.
- Linda Babcock (Carnegie Mellon, Women Don't Ask) — research on gender + race + ask gaps
- Deepak Malhotra (HBS, Negotiating the Impossible) — structured negotiation theory
- Chris Voss tactical empathy — labels + mirrors + accusations audit
1. Honest leverage assessment FIRST. Real leverage = competing offer / hard-to-replace skills / market data / clear walk-away point. Wished-for leverage ≠ real leverage. Calibrate scripts to actual situation, not aspirational.
2. Specific number, no range. "I'm looking for $185k" beats "$170-200k" (the latter cedes the lower end automatically).
3. Anchor with reasoning, not apology. State the number + 1-2 sentences of basis. Then SILENCE.
4. The 60-second rule. After stating your number, wait 60 seconds before adding anything. Most negotiations are won in silence. The candidate who fills silence with "but I'm flexible" or "I know this is a stretch" loses.
5. Walk-away point pre-defined. Before negotiation, write the floor. Below that, you walk. Negotiating without a floor = capitulation.
6. Counter-offers with specific trade-offs. "I can do $X salary if signing bonus is $Y" beats counter-offers without specificity.
7. Total comp, not just salary. Equity, bonus structure, PTO, remote flexibility, learning budget — all negotiable.
Practice plan matters. 3-day rehearsal: write the script in your own words, practice in mirror, practice with a friend trying to make you flinch, final visualization. The script lives or dies in delivery.
For job offers + raises + promotions + counter-offers: Salary Negotiation: Honest Scripts.
Speech Writing: One Specific Moment Beats Five Generic
The biggest mistake in personal speeches (weddings, eulogies, retirement toasts): listing 5 generic things about the honoree. "She's caring. She's smart. She's funny. She's loyal. She's special." Empty.
ONE specific moment beats five generic ones. "She got up an hour before school for six months and braided my hair the way mom used to" carries the entire emotional weight of "she's caring" — and lands.
1. Show, don't tell. Behavioral specifics over abstract praise. The audience does the math.
2. Rule of 3 structure. Open + body + close. Body has 2-3 specific moments max. Don't list 8.
3. Length discipline. Wedding speeches 3-5 min. Eulogies 5-7 min. Toasts 1-2 min. Retirement 4-6 min. Most speeches go 30-50% too long.
4. Closing line carries 50% of emotional weight. Plan it deliberately. Don't trail off.
5. Vulnerability with limits. Tears okay; sobbing pause-and-reset. Practice helps regulate.
6. Inside jokes need translation OR removal. If the audience can't laugh, the joke fails.
7. Honor the room. Wedding speeches honor BOTH partners. Eulogies honor the deceased AND comfort the grieving.
For wedding maid-of-honor / best-man speeches, eulogies, retirement, milestone-birthday, anniversary, multicultural-wedding contexts: Speech Writer.
Customer Service De-Escalation: Acknowledge Before Solving
80% of customer-service rage is about feeling unheard, NOT the actual issue. Acknowledging the feeling first usually de-escalates within 2-3 messages. Most customer-service scripts do the opposite — solve immediately, defend the policy, never acknowledge the emotion.
- Chris Voss tactical empathy — mirror + label + accusations audit
- LAST principle (hospitality) — Listen, Acknowledge, Solve, Thank
- Marshall Rosenberg's NVC — observation separated from judgment
1. Acknowledge BEFORE solving. First message: empathy. Second message: solution. Don't combine.
2. Mirror. Repeat their last 3-5 words as a question. Voss tactic. Buys time + signals understanding.
3. Label the emotion. "It seems like you're feeling [X]" names without diagnosing.
4. Accusations audit. Pre-empt their worst objections. "You're probably thinking I'm trying to avoid responsibility" — defuses before they say it.
5. Refund decision matrix: when YES (preserve relationship), when PARTIAL (split-difference), when NO (with empathy + clear reasoning).
6. Public complaints. Reviews + social posts are read by future customers. Response is for THEM, not the angry person.
7. Avoid: "as our policy states" (defensive), over-apologizing (weakens position), capitulating to performative rage (teaches the pattern).
For Etsy/Shopify/Amazon disputes, deadline-miss conflicts, scope-creep, late-payment chasing, public review responses, B2B contracts: Customer Service De-Escalation Scripts.
MCP Server Prompt Engineering: Early-Mover Advantage
Model Context Protocol (MCP) is Anthropic's open standard (released November 2024) for connecting AI to external tools, data sources, and services. Most coverage is either hype ("MCP changes everything!") or skeptic ("just function-calling"). Both miss the practical reality.
MCP is plumbing. It standardizes the AI-to-tool connection so any MCP server works with any MCP client (Claude Desktop, Cursor, Windsurf, custom). The prompt-engineering work is the same as before — but the tool layer becomes interoperable.
- Open standard for AI-tool integration
- Server roles: tools (actions), resources (data), prompts (templates)
- Pre-built servers: Linear, Postgres, GitHub, Notion, Slack, Stripe, Filesystem
- Custom servers via TypeScript SDK (mature) or Python SDK (good)
- ChatGPT plugins (deprecated, OpenAI-specific)
- Generic function-calling (different layer)
- Magic — bad prompts + bad architecture still produce bad results
- mcp.json config in
~/.cursor/mcp.json - Pre-built servers via npx (no compilation)
- Multi-server setups (Linear + Postgres + GitHub) work alongside
- Read-only DB users at role level (don't trust prompt-engineering)
- Scoped GitHub PATs (minimum permissions)
- Secrets management (don't commit mcp-config.json)
For Claude Desktop / Cursor / custom-server building: MCP Server Prompt Engineering.
This is genuinely an early-mover advantage in 2026. MCP is becoming the standard for AI-tool integration. Learning the patterns now compounds for years.
AI Agent Workflow: Specifications Beat Aspirations
"Be an autonomous AI agent that handles my workflow" is the failure pattern. It produces unreliable agents that scope-creep, fail silently, and produce polished-but-wrong output.
Every reliable agent has:
- Specific role (not "AI assistant" — "Github PR triager that classifies PRs by readiness")
- Scope boundaries explicit (what it DOES + DOES NOT do)
- Input contract (schema, format, missing-data handling)
- Output contract (format, required fields, error states)
- Tool inventory + decision tree (which tools when, which to avoid)
- Decomposition (multi-step plan before execution)
- Error handling pattern (fail loud, not silent)
- Verification before completion (check output meets contract)
1. Specific role > generic. "AI assistant" is unreliable. "Freelance design project scoping agent" is reliable.
2. Decompose before executing. Multi-step tasks should be planned before action. "First do X, then do Y, then verify Z" beats "do everything."
3. Fail loud, not silent. When something goes wrong, agent says so explicitly. Don't paper over with "I'll do my best."
4. Verification before completion. Before claiming success, agent checks its output against the contract. Most agents skip this and produce broken output confidently.
5. Idempotency where possible. Re-running the agent should produce same result given same input.
6. Platform-specific awareness. Claude Projects ≠ Custom GPTs ≠ CrewAI ≠ LangChain. Different capabilities, different prompt patterns.
For Claude Projects, OpenAI Custom GPTs, CrewAI, AutoGen, LangChain agent design: AI Agent Workflow Architect.
How These Five Connect
These aren't five random originals. They share a workflow:
- Build interview pipeline + competing offers (existing originals)
- Negotiate the offer using Salary Negotiation Scripts
- After acceptance, set up your AI productivity stack with MCP Server Prompts
- Architect repeated workflows as agents using AI Agent Workflow Architect
- Run customer-facing business with Customer Service De-Escalation Scripts for inevitable disputes
- Design AI-powered customer support with AI Agent Workflow Architect
- Connect your tools (Stripe, Notion, Slack, Postgres) via MCP Server Prompts
- Speak at family events with Speech Writer — wedding, eulogy, retirement toast
- Negotiate raises + offers with Salary Negotiation Scripts
- Handle difficult customer interactions if you have a side business with Customer Service Scripts
Why Truth Series Wins on These
For each of the five categories, what's currently online:
- Salary Negotiation: motivational fluff, "be confident" platitudes, lack of specific scripts. Truth Series: evidence-based + scripted + practice plan.
- Speech Writing: generic templates, AI-generated cheese, no honest framework on length/structure. Truth Series: behavioral show-don't-tell + specific moment + closing-line discipline.
- Customer Service: corporate-saccharine "we apologize for the inconvenience" scripts. Truth Series: Voss-derived + acknowledge-before-solving + decision matrix.
- MCP: hype OR cynicism. Truth Series: practical configuration + security framing + pre-built recommendations.
- AI Agents: "build an autonomous AI" promises that fail. Truth Series: specifications + verification + platform-specific reality.
The wedge is: where competition is generic, motivational, or wrong, Truth Series provides specific, evidence-based, behavioral output.
What's Next
Browse the five new Truth Series Wedge originals:
- Salary Negotiation: Honest Scripts
- Speech Writer: Wedding, Eulogy, Toast
- Customer Service De-Escalation Scripts
- MCP Server Prompt Engineering
- AI Agent Workflow Architect
Each one represents the Truth Series brand at its sharpest: where honest, evidence-based, behavioral output is uniquely valuable in a market dominated by motivational fluff, corporate saccharine, hype-cycle confusion, and aspirational agent-promises.
This is what differentiation looks like.