Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly read-only and purpose-aligned, but its fallback paths can broaden results and show unverified wallets in a way that may mislead users relying on whale/smart-money signals.
Before installing, confirm you are comfortable with a read-only Polymarket analytics skill that may automatically loosen thresholds or include unverified wallets during data outages. Treat any Unclassified results as raw diagnostics, not verified whale activity.
The skill mixes fixed trigger words with open-ended auto-trigger examples, but it does not define clear precedence or boundaries for activation. This ambiguity increases the chance the skill will run in contexts where the user intended a generic trading discussion rather than a Polymarket whale query, expanding the blast radius of mistaken tool use.
The skill mixes fixed trigger words with open-ended auto-trigger examples, but it does not define clear precedence or boundaries for activation. This ambiguity increases the chance the skill will run in contexts where the user intended a generic trading discussion rather than a Polymarket whale query, expanding the blast radius of mistaken tool use.
Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.
## Error Handling
| Scenario | Resolution |
| -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------- |
| MCP query timeout | Reduce time window to 12h and retry once; if still failing, report error and exit |
| No orders >= $5k in 24h | Lower threshold to $2,000 and retry; note the adjusted threshold in output |
The instruction to lower the threshold from $5,000 to $2,000 when no qualifying orders are found changes the meaning of the skill from large-order monitoring to broader trade monitoring. This can produce materially misleading results for users who rely on the skill to identify whales, especially because the downgrade occurs automatically in an error-handling path rather than by explicit user request.
The error-handling path explicitly says to skip profile filtering and show raw data as "Unclassified" when classification data is empty, which contradicts the skill’s core requirement to show only verified smart-money whale addresses. In practice this can cause the agent to present unverified wallets as whale activity, weakening trust boundaries and enabling misleading or unsafe output during backend degradation.
No suspicious patterns detected.