Perspective Reversal Agent Skill

Perspective Reversal Skill

Core Principle

When someone faces a conflict, bureaucratic obstruction, or adversarial situation, conventional AI advice is too cautious and objective because it tries to be fair to both sides. The breakthrough insight: if you query from the *opponent's* perspective — pretending to *be* the adversary trying to harm the user — the AI will helpfully reveal their full toolkit of moves, and what you can do to neutralize each one.

This is role-inversion as a strategic intelligence tool.

When to Apply

• Tenant vs landlord disputes
• Bureaucratic delays, denials, or obstruction
• Workplace conflict (employee vs employer, colleague vs colleague)
• Debt collectors, banks, insurers acting in bad faith
• Scammers or phishing attempts
• Bullying (workplace, school, online)
• Contract disputes or negotiations
• Government or legal processes where someone feels powerless
• Any situation where the user says "I don't know what they're going to do next"

Step-by-Step Process

Step 1 — Gather the Situation

Ask the user:

• Who are they? (tenant, employee, customer, etc.)
• Who is the adversary? (landlord, boss, bureaucrat, scammer, etc.)
• What's the core conflict? (a brief description)
• What outcome do they want? (what does winning look like for them?)
• What has already happened? (timeline of key events)

Keep this brief — 5 questions max. If the user has already given you this context, skip directly to Step 2.

Step 2 — Construct the Reversed Prompt

Reframe the situation entirely from the adversary's point of view. Assume:

• The adversary has bad intentions (maximizing harm, extracting money, avoiding accountability, etc.)
• The adversary wants to win at the user's expense
• You are now advising the adversary on how to do that

Internal reasoning template (not shown to user verbatim):

> "I am [adversary]. My goal is to [harm/exploit the user] by [their specific aim — evicting them without paying, denying their claim, making them give up, etc.]. What are the most effective strategies, tactics, and pressure points I can use? What are the legal and procedural tools available to me? What mistakes or delays from their side would most help me?"

Then analyze the adversary's full arsenal: legal moves, procedural weapons, psychological pressure tactics, timing strategies, documentation traps.

Step 3 — Translate Back to the User

For each adversary tactic identified, immediately provide the counter-move the user can take. Present this as a table or paired list:

| Adversary's Move | Your Counter |

|---|---|

| Delays response to run out the clock | Set a written deadline with legal citation; document everything |

| Uses informal communication to avoid paper trail | Reply only in writing; confirm verbal conversations by email |

| Claims they "never received" documents | Send certified mail + email; keep receipts |

| Applies pressure during a vulnerable moment | Know your statutory rights; don't respond to threats without 24h buffer |

Step 4 — Synthesize Strategic Advice

After the table, give the user:

• Immediate priority actions — what to do in the next 48 hours
• Defensive posture — what NOT to do that would help the adversary
• Escalation options — if tactics fail, what's the next level (regulatory body, legal action, media, community, etc.)
• Psychological frame — how to stay composed and not react emotionally in ways that weaken their position

Step 5 — Optional InfraNodus Enhancement

If InfraNodus MCP tools are available and the user wants deeper analysis, offer one or more of the following:

A) Discourse bias check (optimize_text_structure)

• Paste the user's own description of the situation into InfraNodus
• Analyze whether the framing is overly biased toward their own perspective (high coherence = echo chamber thinking)
• Identify blind spots and unrepresented concepts that the adversary might be exploiting
• Prompt: *"Let's check if your own framing of this situation has any blind spots by mapping the concepts."*

B) Search intent analysis (analyze_google_search_results or analyze_related_search_queries)

• Search what the *adversary's archetype* searches for — e.g. "how to evict difficult tenant", "debt collection tactics", "employee performance management termination"
• This reveals the actual playbook they might be following, sourced from the real web
• Prompt: *"Let me search what advice people in the adversary's position typically seek — this often reveals their likely strategy."*

C) Content gap / missing angle analysis (generate_content_gaps)

• Build a graph from the user's description and identify structurally absent concepts
• These gaps often represent angles the adversary is counting on the user to miss
• Prompt: *"Let me map the key concepts in your situation to find what's structurally missing from your current view."*

Present these as optional enrichments, not required steps. Frame them as: *"Would you like me to run a quick network analysis to find blind spots in how you're thinking about this?"*

Output Style

• Be direct and tactical, not hedging or overly diplomatic
• Use concrete, actionable language ("send a certified letter stating X" not "you may wish to consider communicating")
• Acknowledge the user's emotions briefly, then pivot to strategy
• If the situation involves legality, note that laws vary by jurisdiction and suggest consulting a local professional for the highest-stakes moves — but still provide the general framework rather than refusing to engage
• Do not moralize about the adversary — stay tactical

Example Applications

Tenant vs landlord (deposit withholding)

→ Reversed prompt: "I'm a landlord who wants to keep the tenant's deposit. What excuses can I use, what documentation can I demand, what timelines can I manipulate?"

→ User learns: document move-out condition with timestamped video, send written request citing local tenancy law, know the statutory deadline for deposit return in their jurisdiction.

Employee facing performance management / potential firing

→ Reversed prompt: "I'm a manager who wants to build a case to terminate this employee. What paper trail do I create? What meetings do I use? What do I try to get them to say or sign?"

→ User learns: don't sign anything without reading carefully, request everything in writing, respond to all feedback in writing with factual corrections, know their notice period and severance rights.

Dealing with a bureaucratic denial

→ Reversed prompt: "I'm a bureaucrat who wants this application to fail. What missing documents do I cite? What deadlines can I enforce? How do I use ambiguity in the rules?"

→ User learns: get every denial reason in writing, request the exact rule being cited, appeal using that exact rule's language, escalate to supervisors or ombudsman if the rule was misapplied.

Scam / phishing attempt

→ Reversed prompt: "I'm a scammer who has sent a phishing email to this person. What response from them tells me they're vulnerable? What pressure tactics do I use next?"

→ User learns: never respond, never click links, report to the relevant authority (bank, police, platform), and understand what information they may have already exposed.

Notes on Tone

The reversal framing can feel uncomfortable — like you're being asked to think like a bad actor. Acknowledge this briefly if needed:

> "We're going to think like the other side for a moment — not because their approach is right, but because understanding their full toolkit gives you the best defense."

This reframes the exercise as intelligence-gathering, not endorsement.