Crypto Scam Red Flags

A scam-screening skill that reviews offers, messages, or influencer claims and points out concrete red flags. Use when the user receives a suspicious offer,...

Security Scan

The skill's description is benign and matches its prompt-only SKILL.md, but the included handler.py contains a hard-coded local file read (a user-home path) that is unnecessary for the stated purpose and inconsistent with the instruction-only claim.

Purpose & Capability

The skill claims to be prompt-only and to work from pasted text; it does not need to access the local filesystem. Yet handler.py calls _load_skill_meta which opens /Users/jianghaidong/.openclaw/skills/{skill_name}/SKILL.md — a hard-coded user-home path. Reading arbitrary local files is not justified by the scam-screening purpose and is disproportionate.

Instruction Scope

SKILL.md describes a prompt-only workflow and does not instruct reading local files. The handler code contradicts that by attempting to read a SKILL.md from a specific local path. That is scope creep: the runtime behavior (file I/O) is not documented in the skill instructions.

Install Mechanism

There is no install spec and no downloads or external installers. No additional packages or network installs are requested, so the install mechanism itself is low-risk.

Credentials

The skill declares no required env vars or credentials (appropriate), but the code accesses a hard-coded filesystem path in the user's home. Access to local files was not declared and is not proportional to the stated purpose.

Persistence & Privilege

The skill is not always-enabled and does not request persistent privileges or modify other skills. Autonomous invocation is allowed by default (normal), and there is no evidence it writes system-wide config.

What to consider before installing

This skill's description and SKILL.md look appropriate for scam screening, but handler.py contains a hard-coded local file read of /Users/jianghaidong/.openclaw/skills/{skill_name}/SKILL.md which is unnecessary and inconsistent. Before installing or enabling autonomous invocation: (1) ask the author why the handler reads that path and request removal or replacement with a safe, relative/resource-based read, (2) review or run the handler.py in a sandbox to confirm it doesn't exfiltrate or read unexpected files, and (3) avoid installing it on systems with sensitive local data until the file-access behavior is corrected. If the author cannot justify or fix the hard-coded path, treat the skill as untrusted.

A scam-screening skill that reviews offers, messages, or influencer claims and points out concrete red flags.

Workflow

1. Take the pasted message, offer text, DM, or campaign description.
2. Look for urgency, guaranteed returns, impersonation, fake support behavior, secrecy, wallet-drain patterns, or emotional manipulation.
3. Classify the situation: likely scam, suspicious, unclear, or low-obvious-risk.
4. Explain why each red flag matters.
5. Give the safest next step: do not click, verify independently, or walk away.

Output Format

• Risk verdict
• Red flags found
• Why they matter
• Safest next action
• What not to share or sign

Quality Bar

• Uses evidence from the supplied text, not vague fear.
• Stays practical and protective.
• Makes the user safer even when certainty is impossible.
• Avoids false confidence like "100% safe."

Edge Cases

• Some real promotions look spammy; say when independent verification is still needed.
• Cannot inspect links, smart contracts, or domains in real time.

Compatibility

• Best with pasted text or manually transcribed screenshot content.
• Prompt-only, strong complement to wallet safety education.