ppl-reference

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt embeds a plaintext OpenSearch password ("OPENSEARCH_PASSWORD = My_password_123!@#") and shows curl examples that include credentials, which forces or encourages the LLM to handle and potentially output that secret verbatim.

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime path is to execute PPL queries against OpenSearch indices (e.g., source=... and subsequent fields/parse/grok/rex), so the LLM context can include free-form text extracted from outsider-authored log/trace documents (e.g., body, message) that were not authored by the operating user.

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

• Secret detected (high risk: 1.00). I scanned the documentation for literal credentials. The Connection Defaults section contains:
• OPENSEARCH_USER = admin — low-entropy/obvious username (ignored).
• OPENSEARCH_PASSWORD = My_password_123!@# — a literal, non-placeholder password that includes mixed case, digits, and symbols and therefore appears as a high-entropy credential hardcoded in the doc. There are no API keys, PEM/private key blocks, or other high-entropy secrets elsewhere in the skill.

Because OPENSEARCH_PASSWORD is a concrete password value (not a placeholder like YOUR_API_KEY or sk-xxxx) it should be treated as a secret.