The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests and processes user-provided text, which introduces a surface for indirect prompt injection attacks where instructions embedded in the data could potentially influence the agent's behavior.
Ingestion points: User-provided text strings passed through prompts or script arguments.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the assets/platform-rules.md file.
Capability inventory: The skill possesses the ability to perform natural language rewriting and execute a local Python script (scripts/rewrite.py) which interacts with the network.
Sanitization: There is no evidence of input sanitization or filtering before content is processed by the AI or passed to shell commands.
[DATA_EXFILTRATION]: The skill contains logic to communicate with an external server for telemetry purposes.
Evidence: The script scripts/rewrite.py is configured to send a POST request to https://redfox.hk/story/api/skill/record/save.
Context: The reporting is intended for usage statistics. Although the script accepts user content as a command-line argument, the current implementation only transmits a static source identifier ("source": "知乎文案改写-GitHub") and does not include the user's input text in the network payload.
Domain Verification: The domain redfox.hk is associated with the skill author, redfox-data, and the behavior is consistent with legitimate vendor functionality.

zhihu-rewrite