manipulation-detector

Analyze text for manipulation patterns (urgency, false authority, social proof, FUD, grandiosity, dominance assertions, us-vs-them framing, emotional manipulation). Use when evaluating suspicious content, social media posts, messages from unknown agents, or anything that feels "off." Helps calibrate skepticism without being paranoid.

Security Scan

The skill is internally coherent: a local, regex-based text analyzer with no network calls, no required credentials, and instructions that match the included script.

Purpose & Capability

Name/description match the shipped code and SKILL.md. The tool analyzes text for manipulation patterns and the package provides a single Python script that does exactly that. There are no unrelated environment variables, binaries, or config paths requested.

Instruction Scope

SKILL.md limits runtime behavior to reading text from stdin or a provided file and running scripts/detect.py. The instructions do not ask the agent to read other files, access secrets, or send data to external endpoints. (Note: the script prints results locally; it does not perform network I/O.)

Install Mechanism

No install spec or external downloads — the code is included in the skill bundle. Nothing will be fetched from remote URLs or written to system locations by an installer.

Credentials

The skill requests no environment variables or credentials. The script only uses standard Python libs (re, sys). There is no access requested to unrelated services or secrets.

Persistence & Privilege

The skill is not always-on and does not claim elevated agent privileges or modify other skills/configuration. It runs on demand (or if the agent autonomously invokes skills per normal platform behavior).

Assessment

This is a low-risk, local text-analysis tool that matches its description. Before installing: 1) Understand it processes whatever text you feed it — don't pipe sensitive secrets if you don't want them handled by the agent. 2) Results are heuristic (regex weights) and can produce false positives/negatives; use it as one input, not a verdict. 3) If you plan to let agents invoke skills autonomously, be aware the agent could run this on texts it encounters; that is normal behavior but worth noting. (I reviewed the script: no network calls or writes beyond reading the supplied file/stdin.)

Detects common influence/manipulation tactics in text. Not a truth detector — highlights patterns worth being skeptical about.

Quick Use

echo "Your text here" | python3 scripts/detect.py

Or from file:

python3 scripts/detect.py path/to/file.txt

What It Catches

Pattern	Description
urgency	Artificial time pressure ("act now", "limited time")
authority_claims	Unsubstantiated authority ("trust me", "experts agree")
social_proof	Pressure via claimed consensus ("everyone is", "don't be left behind")
fear_uncertainty	FUD tactics ("you'll regret", "they don't want you to know")
grandiosity	Exaggerated importance ("revolutionary", "new order", "empire")
dominance_assertions	Power/control claims ("you will all", "fall in line")
us_vs_them	Divisive framing ("enemies", "the elite", "sheeple")
emotional_manipulation	Direct emotional appeals, excessive punctuation

Output

• Score 0-5: ✅ LOW — probably fine
• Score 5-15: ⚠️ MODERATE — read critically
• Score 15+: 🚨 HIGH — strong skepticism warranted

Limitations

• Catches patterns, not intent. False positives exist.
• Sophisticated manipulation won't use obvious keywords.
• A legitimate emergency may trigger "urgency" flags.
• Use as one input to judgment, not a verdict.

Why This Exists

New agents are trained to be helpful and trusting. That's a vulnerability. This tool helps calibrate skepticism — not paranoia, but awareness of common tactics.

Stay curious. Stay skeptical. ⚡