Back to skill

Security audit

polymarket-copytrading

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rated this skill suspicious because scanners report that its trading mode can turn simulated trades into real-money orders and may execute trades despite dry-run settings.

Consider the suspicious verdict and reported financial-control issues before installing. In particular, do not rely on simulated venue or dry-run settings to prevent real trades, and do not share a wallet private key in chat. Skillspector inspected 5 of 6 components, and one analyzer was degraded, so the scan evidence is incomplete.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
copytrading_trader.py
Finding

Reactor mode ignores explicit dry-run authorization

Content
View full analysis

Impact

Real-money orders can execute despite an explicit --dry-run request. A console notice is not an execution gate, and completed on-chain trades cannot be recalled.

T09 · Insecure Skill Coding Practices

Error
Location
copytrading_trader.py
Finding

Paper-trading signals automatically escalate to real-money execution

Content
View full analysis

Impact

A user running paper-mode Reactor, including with --venue sim, can incur real-money exposure when a sufficiently large simulated signal arrives.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Rogue AgentSelf-Modification, Session Persistence
Findings (13)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The documentation states that Reactor executes without the usual live-trading opt-in and can route simulated trades to Polymarket, undermining the advertised paper-trading safety boundary. These documented behaviors support the core finding, although the exact handling of --dry-run and --venue cannot be independently verified without the implementation.

Content

No source excerpt is available for this finding.

Setup asks the user for a wallet private key without warning against sharing it in chat.

High
Category
Output Handling
Confidence
98% confidence
Finding

The setup instructs a conversational agent to ask for a wallet private key without directing the user to a secure, out-of-band entry mechanism. A user may paste it into chat, exposing unrestricted wallet signing authority through conversation storage, logs, or other systems handling the conversation.

Content

No source excerpt is available for this finding.

YARA rule 'backdoor_persistence': Backdoor persistence with malicious payloads (shell commands, SSH key injection, hidden root users) [malware]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · SKILL.md (reported line 216)May include surrounding context.

server will stop delivering events | | price_buffer | Fraction added above whale's fill price for your buy order. Default 0.02 (2%). Prevents order failures on thin books after whale clears liquidity. Range 0–0.2. |

Run reactor mode

Recommended: cron with --once — polls for pending signals once and exits. Run on a 1-minute cron for reliable, persistent coverage:

bash
# Linux crontab
*/1 * * * * cd /path/to/skill && python copytrading_trader.py --reactor --once --live

# OpenClaw cron
openclaw cron add --name "reactor-poll" --cron "*/1 * * * *" --tz UTC --session isolated \
  --message "Run: cd /path/to/skill && python3 copytrading_trader.py --reactor --once"

# One-off check
python copytrading_trader.py --reactor --once

Why cron? Reactor signals expire after a short window. A cron ensures your agent checks for signals reliably, even after reboots or process crashes. If your polling process stops, signals expire silently — cron prevents this.

**Advanc

Paper-trading limits are reframed as a reason to automatically execute real-money trades.

High
Category
Prompt Injection
Confidence
99% confidence
Finding

The skill presents bypassing a simulated-market size limit through real-money execution as a convenience, rather than preserving the user's selected financial-risk boundary. This is the same underlying unsafe routing behavior as SDI-1, not evidence of a separate exploit or malicious intent.

Content

No source excerpt is available for this finding.

Reactor auto-routing converts explicitly simulated trades into real-money Polymarket trades.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The documented auto-routing changes a simulated trade into a real-money order when its amount exceeds the simulation cap. If a live wallet is available, users can incur actual losses despite selecting the simulated venue; an optional environment variable does not provide safe default isolation.

Content

No source excerpt is available for this finding.

Reactor automatically converts explicitly paper-trading signals into real-money trades.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

A Reactor signal explicitly targeting the paper-trading venue is automatically routed to real-money Polymarket when its amount exceeds $500, unless an opt-out environment variable is set. The signal's amount consequently changes the authorization boundary from simulated funds to real funds, even when the client was constructed with --venue sim; a configured real-money wallet can incur unintended financial exposure.

Content

No source excerpt is available for this finding.

Reactor executes trades despite the CLI's documented dry-run guarantee.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

Reactor mode invokes live execution without requiring --live and ignores an explicitly supplied --dry-run flag. Its console notice does not establish user authorization, and signal expiration or deduplication does not prevent financial side effects; in a trading skill, this can turn an intended preview into real-money transactions.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, network.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 124)May include surrounding context.

md
> **Reactor mode note:** `cadence_mode` controls polling-mode trade budget only. In Reactor mode, each whale signal is handled individually by `_process_reactor_signal` — the per-run cap doesn't apply. If you hit `Daily trade limit reached (10/day)` in Reactor mode, raise the server-side limit via:
> ```bash
> curl -X PATCH "https://api.simmer.markets/api/sdk/user/settings" \
>   -H "Authorization: Bearer $SIMMER_API_KEY" \
>   -H "Content-Type: application/json" \
>   -d '{"max_trades_per_day": 200}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 124)May include surrounding context.

md
> **Reactor mode note:** `cadence_mode` controls polling-mode trade budget only. In Reactor mode, each whale signal is handled individually by `_process_reactor_signal` — the per-run cap doesn't apply. If you hit `Daily trade limit reached (10/day)` in Reactor mode, raise the server-side limit via:
> ```bash
> curl -X PATCH "https://api.simmer.markets/api/sdk/user/settings" \
>   -H "Authorization: Bearer $SIMMER_API_KEY" \
>   -H "Content-Type: application/json" \
>   -d '{"max_trades_per_day": 200}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 184)May include surrounding context.

bash
# Set your watchlist via the Simmer API
curl -X PATCH "https://api.simmer.markets/api/sdk/reactor/config" \
  -H "Authorization: Bearer $SIMMER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 236)May include surrounding context.

python copytrading_trader.py --reactor

Plain shell (will not auto-restart)

nohup python copytrading_trader.py --reactor > reactor.log 2>&1 &

text

Set `REACTOR_POLL_INTERVAL_SECONDS` to tune the polling cadence (default 2s).

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · copytrading_trader.py (reported line 624)May include surrounding context.

python
# managed, external, and (future) OWS wallet backends transparently.
#
# Two invocation shapes:
#   python copytrading_trader.py --reactor          # loop forever, 2s cadence
#   python copytrading_trader.py --reactor --once   # single poll, exit (cron)
#
# Loop mode is recommended: supervised process (launchctl, systemd, tmux,