media-gpac

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt shows and requires embedding encryption keys/KIDs directly in CLI arguments and the CENC XML (and examples for decryption), which forces an agent to output secret values verbatim when performing encrypt/decrypt operations.

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

• Secret detected (high risk: 1.00). I scanned the skill for literal, high-entropy values that could be used to access or decrypt content. I flagged values that are 16-byte (32-hex-char) keys/KIDs used as ClearKey content keys — these are high-entropy, valid-format crypto keys and would be usable to decrypt CENC content if actually used.

Flagged items (present as literal values in the docs / examples / scripts):

• Quick start / helper invocation:
• --key-id 0123456789abcdef0123456789abcdef
• --key fedcba9876543210fedcba9876543210
• Example drm.xml (Example 3):
• KID="0xABCDEF01234567890ABCDEF012345678"
• value="0x112233445566778899AABBCCDDEEFF00"
• references/gpac.md and other examples repeat similar 0x... KID/value examples (e.g., 0x112233445566778899AABBCCDDEEFF00 and first_IV="0x0123456789abcdef").
• scripts/gpac.py: the DRM-XML writer uses a default IV "0x0123456789abcdef" and writes KID/key values verbatim into generated XML (so supplying literal key/KID args would embed them).

Why these are flagged:

• They are high-entropy hex strings of the correct lengths for CENC keys/KIDs (16 bytes / 32 hex chars) and would be functional decryption keys, not mere low-entropy placeholders.
• They are literal values in the repository text and script examples (not redacted or truncated), so they constitute exposed credentials.

Notes on items I did NOT flag:

• The example IV "0x0123456789abcdef" is an IV/nonce (not a secret that grants access) — I mention it but it is lower sensitivity compared to keys.
• Environment variable names, command names, and obvious placeholders (none present here) were ignored per the rules.

Conclusion: the doc contains literal, usable ClearKey content keys / KIDs embedded in examples and scripts, which meet the definition of secrets.