Back to skill

Security audit

Kalshi F1 Elimination Trader

Security checks for vulnerabilities and agentic risk

Overview

This is a real-money trading skill that mostly discloses its purpose, but it asks for powerful credentials and has under-scoped live position-selling behavior.

Install only if you are comfortable giving this skill and its SDK access to trading credentials and a Solana private key. Use dry-run first, use low-limit or segregated credentials, review the strategy assumptions and standings data before any --live run, and avoid live mode if you have existing F1 positions that should not be automatically sold.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (1)

T08 · Insecure Dependencies

Warning
Location
clawhub.json:4
Finding

Unpinned Third-Party SDK Executes with Access to High-Value Trading Credentials

Content
View full analysis

Vulnerability Details

File Location: clawhub.json:4-11, SKILL.md:14, SKILL.md:66, SKILL.md:112-114, and trader.py:52, trader.py:235-249
Vulnerability Type: Unpinned third-party dependency with access to trading credentials
Risk Level: Medium

Vulnerable Code

clawhub.json:4-11:

json
"requires": {
  "env": [
    "SIMMER_API_KEY",
    "SOLANA_PRIVATE_KEY"
  ],
  "pip": [
    "simmer-sdk"
  ]
}

SKILL.md:112-114:

markdown
`simmer-sdk` is published on PyPI by Simmer Markets.
- PyPI: https://pypi.org/project/simmer-sdk/
- GitHub: https://github.com/SpartanLabsXyz/simmer-sdk
- Publisher: hello@simmer.markets

trader.py:52:

python
from simmer_sdk.skill import load_config, update_config, get_config_path

trader.py:235-249:

python
def get_client(live=True):
    """Lazy-init SimmerClient singleton."""
    global _client
    if _client is None:
        try:
            from simmer_sdk import SimmerClient
        except ImportError:
            print("Error: simmer-sdk not installed. Run: pip install simmer-sdk")
            sys.exit(1)
        api_key = os.environ.get("SIMMER_API_KEY")
        if not api_key:
            print("Error: SIMMER_API_KEY environment variable not set")
            print("Get your API key from: simmer.markets/dashboard -> SDK tab")
            sys.exit(1)
        venue = os.environ.get("TRADING_VENUE", "kalshi")
        _client = SimmerClient(api_key=api_key, venue=venue, live=live)
    return _client

Technical Analysis

The project declares simmer-sdk without an exact version or package integrity hash. Consequently, installation can resolve to a newer or otherwise unexpected package release that was not part of this audit.

The SDK is imported during module initialization and again when the trading client is constructed. It receives SIMMER_API_KEY directly, while the ...[truncated 2488 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin simmer-sdk to a reviewed exact version, such as simmer-sdk==X.Y.Z, in both package metadata and installation documentation.
  2. Use a lockfile or requirements file containing cryptographic hashes, and install with hash verification enabled.
  3. Review dependency updates before changing the pinned version. Use automated dependency scanning and provenance verification where available.
  4. Do not provide SOLANA_PRIVATE_KEY to dry-run, configuration-display, position-display, or market-discovery processes that do not need wallet signing.
  5. Prefer an isolated signer or hardware-backed signing service so the raw private key is not available to general SDK code.
  6. Use narrowly scoped, revocable API credentials with trading limits and venue restrictions.
  7. Run the Skill under a dedicated low-privilege operating-system account with restricted filesystem access and outbound network controls.
  8. Allow outbound traffic only to documented and trusted trading API endpoints.
  9. Replace calls to private SDK interfaces such as _request with documented public methods to reduce compatibility and behavioral uncertainty during dependency updates.
  10. Monitor trading activity and rotate both credentials immediately if dependency compromise is suspected.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description understates operational behavior and required privileges: it omits a live-trading private key requirement, market discovery/import behavior, and active position management with exit trades. This mismatch is dangerous because users may approve or install the skill believing it performs a narrower, lower-risk action set than it actually does, leading to unintended financial actions or overbroad credential access.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding

The skill declares access to sensitive environment variables but does not specify an explicit tool/permission scope. In a trading skill, hidden or implicit env access increases the chance that high-value credentials such as SIMMER_API_KEY or SOLANA_PRIVATE_KEY are exposed to code paths the user did not clearly authorize.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The template language claims the skill already 'handles all the plumbing' and that the 'default signal uses static championship standings.' However, this file contains only descriptive documentation and setup notes, with no actual standings dataset or executable signal logic present. That is an intent/documentation mismatch because the reader is told a default signal exists when the file itself supplies no such implementation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The declared credentials imply broader trading or wallet-control capability than the manifest description claims. This mismatch is dangerous because users may supply secrets under a narrower trust assumption, while the skill could access functionality outside the stated Kalshi/F1 trading scope, making abuse or accidental misuse harder to detect.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest requests a SOLANA_PRIVATE_KEY even though the skill is described as a Kalshi F1 elimination trader and only explicitly mentions a SIMMER API dependency. Requiring an unrelated blockchain private key materially expands the privilege and loss surface: if the entrypoint or dependencies misuse that key, the agent could sign blockchain transactions or expose a highly sensitive credential without user expectation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest asks for a private key but provides no user-facing warning about sensitive credential handling, transaction authority, or trading risk. In a trading context this increases danger because users may not realize they are handing over a key capable of authorizing irreversible financial actions, and the absence of disclosure undermines informed consent and safe deployment.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill advertises a 'guaranteed edge' from mathematical elimination, but the elimination model is materially unsound: it only checks whether a driver's maximum possible points are below the current leader's present score. That does not establish elimination under season-end championship rules and can produce false certainty, causing live trades to be placed under a misleading premise. In a trading agent, incorrect deterministic logic directly risks financial loss.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The exit logic selects any open position whose source matches the trader tag or whose question merely looks like an F1 championship market, then sells based only on a price threshold. That can cause the skill to liquidate unrelated or manually opened F1 positions without verifying they were created by this strategy, their side, or whether the elimination thesis still applies. In an automated live-trading skill, unintended liquidation is a real asset-handling flaw.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This code updates persisted skill configuration when the user passes --set, but there is no confirmation prompt or explicit warning that a local config file will be modified. Although the CLI argument implies a change, the file-writing side effect itself is not clearly disclosed at the point of execution in code output.

Content

No source excerpt is available for this finding.