typewriter-video

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

• Potentially malicious external URL detected (high risk: 0.80). The README explicitly instructs an AI agent to clone/install the skill from the remote GitHub repo (https://github.com/yammaku/typewriter-video), which the agent will fetch at runtime and whose repository content directly defines the skill's instructions/behavior — making this a runtime external dependency that can control prompts or execute code.

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

• Secret detected (high risk: 1.00). I reviewed the entire skill docs for high-entropy literal values that look like real credentials.

Flagged item:

• The README contains a join URL with an invitation token: https://member.pathunfold.com/join?invitation_token=8b9f3cfa9991545682f3123b04970a0875311e46-27313521-3b92-472f-ae3f-b970c03f532e This is a long, random-looking token in a join link that likely grants access to a membership space — it meets the "high-entropy value that provides access" definition and is not a placeholder. I therefore treat it as a secret and flag it.

Ignored items (not flagged) and why:

• Public asset identifiers/URLs (GitHub, YouTube, asset GUID like e2319d8e-36b2-43f9-aca2-5aa4d427adf5) — these are public references, not credentials.
• Build/package metadata (package.json/lockfile versions, integrity hashes) — these are checksums and package metadata, not secrets.
• Installation commands, example file paths, font/sound filenames, and numeric constants (e.g., CH3_OFFSET = 228) — these are config/documentation, not credentials.
• No API keys (sk-...), private key PEM blocks, or other obvious token formats were present elsewhere.

Recommendation: remove or redact the invitation_token from public docs (replace with a placeholder like YOUR_INVITE_TOKEN or a redaction marker) if it is meant to be private.