The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to download and run the exportdou package from the official npm registry. It also mentions installation via other platforms like ClawHub and SkillHub.
[COMMAND_EXECUTION]: The skill relies on shell commands to interact with the exportdou CLI for authentication, task management, and data retrieval.
[CREDENTIALS_UNSAFE]: While the skill manages an EXPORTDOU_API_KEY, it follows secure practices by storing the key in a local configuration file (~/.exportdou/config.json) with restricted filesystem permissions (0o600). The instructions explicitly forbid the agent from printing or exposing the API key in its responses.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface as it ingests untrusted data from Douyin comments and share text for analysis.
Ingestion points: Untrusted data enters the agent context through the output of npx exportdou preview and npx exportdou download (SKILL.md).
Boundary markers: The instructions recommend using a small sample (maximum 50 rows) for analysis and caution against placing large exports directly in the model context.
Capability inventory: The CLI tool can write to the local filesystem (--output) and perform network operations to the vendor's API (packages/cli/src/api.ts).
Sanitization: The CLI tool provides structured JSON output, reducing the risk of schema confusion during interpolation.

exportdou