The Agent Skills Directory

[COMMAND_EXECUTION]: The ima_notes_tool.py script utilizes subprocess.run to execute a secondary local script, ima_tool.py, for command orchestration. The execution uses a list of arguments with the current Python interpreter, which is a safe practice to avoid shell injection.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to https://ima.qq.com using the urllib.request library. This is the official endpoint for the IMA OpenAPI, a well-known service from Tencent.
[DATA_EXFILTRATION]: The tool provides functionality to read local files via the --content-file argument in the create-note and append-note commands. This data is then transmitted to the IMA API. This is an intended feature for document import and follows the primary purpose of the skill.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external data.
Ingestion points: Note content is retrieved via the IMA API (read-note) and local files can be read via the --content-file argument.
Boundary markers: The instructions do not define delimiters or warnings for the agent to ignore instructions embedded within note content.
Capability inventory: The skill has the ability to read local files and perform network operations via API calls.
Sanitization: There is no evidence of content sanitization or validation before the data is presented to the agent.
[CREDENTIALS_UNSAFE]: The skill manages API credentials (client_id and api_key) through a configuration file or environment variables. Instructions are provided for the agent to assist in setting up these credentials, which is a standard configuration procedure.

ima-notes