autonomous-agents

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt embeds actual-looking API keys/passwords and maps banned values to their "correct" secret replacements, which implies the agent may need to output or copy those secret strings verbatim (high exfiltration risk).

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

• Secret detected (high risk: 1.00). Flagging logic:
• I only flag literal, high-entropy strings that appear usable as keys/passwords in the doc.

Flagged items (real, high-entropy literals present):

• 6bb5a5733b79dc30f42ea4ab6a95b9a0 — 32-char hex-looking string documented as a Kie.ai key in wrong entity (appears to be an actual API key).
• cb711f74a221be35a20df8e26e722e04 — 32-char hex-looking string documented as the correct SuperSeller key (appears to be an actual API key).
• 7Te1UudOE54aU79xrZ — 19-character mixed-case alphanumeric string documented as an "Old RackNerd password" (appears to be a real password).

Ignored / not flagged (reasons):

• ${REDIS_PASSWORD} — unexpanded env var placeholder (documentation placeholder).
• uad.garage.doors@gmail.com and 1shaifriedman@gmail.com — email addresses, not high-entropy secrets.
• SHAI_OPS_GROUP_ID — environment variable name referenced but no value present.
• internalBoss — simple word / session name (low entropy example).
• interactive_login.js — filename, not a secret.
• Any other context like "YOUR_API_KEY", truncated values, or example strings are treated as placeholders per the documentation rules and were ignored.

Conclusion: There are literal high-entropy credentials present in the BANNED VALUES table (two 32-char hex strings and one high-entropy password).