soma

Fail

Audited by Snyk on May 4, 2026

Risk Level: HIGH
Full Analysis

HIGH W007: Insecure credential handling detected in skill instructions.

  • Insecure credential handling detected (high risk: 0.90). The skill explicitly instructs filling a .env with secrets (SOMA_SECRET_KEY, HF_TOKEN, S3 keys), shows code examples that embed a secret (Keypair.from_secret_key("YOUR_SECRET_KEY")) and guides copying/pasting keys for on-chain operations, which encourages requesting and inserting secret values verbatim into commands or code—creating a high exfiltration risk.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (high risk: 0.90). The skill explicitly streams and ingests public, user-generated datasets (e.g., "The Stack v2" via HuggingFace in references/quickstart-patterns.md and data-strategies.md), queries open targets and model manifests (client.get_targets, get_model_manifests), and downloads third-party model weights and submission data (client.fetch_model, client.fetch_submission_data) which the agent is required to read and use to drive scoring, submission, and training decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

  • Potentially malicious external URL detected (high risk: 0.90). The Getting Started instructions tell users to run "curl -fsSL https://sup.soma.org | bash", which downloads and immediately executes a remote installer script at runtime (https://sup.soma.org) and is presented as a required dependency for the workflow.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

  • Direct money access detected (high risk: 1.00). The skill explicitly integrates with a blockchain wallet and on-chain transaction APIs. It requires a SOMA_SECRET_KEY, shows using soma wallet commands and Keypair.from_secret_key, and contains explicit functions to send transactions: submit_data, create_model (with stake/commission), commit_model/reveal_model (commit-reveal on-chain), claim_rewards, merge_coins, get_balance, and other bond/coin/gas operations. These are direct crypto/blockchain signing and token management actions (moving tokens, staking, claiming rewards), not generic API or browser automation. Therefore it provides direct financial execution authority.

Issues (4)

W007
HIGH

Insecure credential handling detected in skill instructions.

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

W012
MEDIUM

Unverifiable external dependency detected (runtime URL that controls agent).

W009
MEDIUM

Direct money access capability detected (payment gateways, crypto, banking).

Audit Metadata
Risk Level
HIGH
Analyzed
May 4, 2026, 03:01 PM
Issues
4