owasp-security

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The document includes hardcoded API keys, plaintext passwords, base64 secrets, and prompt-injection examples that demonstrate or would cause an agent to output secret values verbatim, which is insecure credential handling.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow is to read the user’s provided code/config/description (outsider-authored free text) and include it in the LLM context for analysis, creating an indirect prompt-injection surface via the user’s message content.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I flagged a literal, long "sk-..." API key present in the repository because it is a hardcoded, high-entropy-looking value used directly in code (usable credential pattern). Specifically, examples/cryptographic-failures.js contains:
• const api_key = "sk-abc123xyz789defgh1234567890"; — a concrete API-key-format string hardcoded in source (VULNERABLE example). This meets the definition of a secret (literal key) and is not a generic placeholder.

I ignored other occurrences that resemble keys for the reasons below:

• Short or clearly example placeholders such as "sk-abc123xyz789" (in SKILL.md and examples/k8s-rbac.yaml) are low-entropy/documentation examples and are treated as placeholders.
• Truncated or ellipsized mentions (e.g., "sk-abc123xyz789...") are redacted/truncated in-context examples and are ignored per the rules on truncated/redacted values.