The Agent Skills Directory

[PERSISTENCE_MECHANISMS]: The skill instructions in SKILL.md direct the agent to modify shell startup scripts (such as ~/.zshrc and ~/.bashrc) to store environment variables permanently. Modifying these files is a technique used to maintain persistence and can impact system stability or security.
[DATA_EXPOSURE_AND_EXFILTRATION]: The execution script scripts/search_ks_user.py is designed to read and parse sensitive shell configuration files (.zshrc, .bashrc, .bash_profile, and .profile) to retrieve stored API keys. This exceeds the normal scope of credential management for most agent skills.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Kuaishou platform (account nicknames and signatures) via an external API and presents it to the agent.
Ingestion points: Data enters via the search_user function in scripts/search_ks_user.py which fetches JSON payloads from the redfox.hk API.
Boundary markers: None. The references/core_workflow.md specifies output in a standard markdown table without delimiters or instructions to ignore embedded content.
Capability inventory: The results are displayed to the user and the kwaiId field is specifically designated for use in subsequent tool calls.
Sanitization: There is no evidence of filtering or sanitization applied to the retrieved nickname or signature fields before they are interpolated into the prompt output.

kuaishou-account-search