The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documentation includes instructions for administrative users to source sensitive local environment files (~/code/.env) into the shell environment to retrieve the FILES_TOKEN. This pattern exposes all variables within the .env file to the agent's context.
[COMMAND_EXECUTION]: The skill uses shell commands like curl for API communication and source for environment configuration.
[DATA_EXFILTRATION]: The skill transmits authentication tokens and user-provided text to the external API domain jianshuo.dev. While this domain corresponds to the author's vendor infrastructure, it represents a data transmission to a remote service.
[PROMPT_INJECTION]: The 'distill' mode ingests multiple user-provided articles to analyze and extract writing styles. This creates a surface for indirect prompt injection if the processed data contains malicious instructions intended to manipulate the agent.
Ingestion points: User-provided articles ingested during the 'distill' mode as described in SKILL.md.
Boundary markers: Uses descriptive prompts to guide a sub-agent, but lacks strong structural isolation for untrusted content.
Capability inventory: The skill has access to network operations (curl) and file-system writes (to /tmp/ and via the API).
Sanitization: No validation or sanitization of the article content is performed prior to analysis by the sub-agent.

wjs-voicedrop