The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the shell_exec tool to perform environment detection, specifically running python -c "import platform; print(platform.system())" during its initialization phase. This establishes a capability that could be targeted for exploitation if the agent is compromised.
[EXTERNAL_DOWNLOADS]: The skill is designed to continuously fetch and analyze data from arbitrary external sources using web_search and web_fetch tools as part of its core intelligence gathering function.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It retrieves content from untrusted external web pages and processes it to extract entities and sentiment without using boundary markers or sanitization.
Ingestion points: External web content retrieved via web_fetch in Phase 3.
Boundary markers: Absent; the agent is instructed to process full content from fetches without specific delimiters to separate untrusted data from instructions.
Capability inventory: shell_exec, file_read, file_write, schedule_create, and web_fetch.
Sanitization: Absent; there is no logic to validate or escape external content before it is processed by the AI.
[DATA_EXFILTRATION]: The skill maintains a combination of file_read and web_fetch tools. This configuration creates a potential path for data exfiltration where malicious instructions from a fetched website could instruct the agent to read local files and send the data to a remote server.

collector-hand-skill