The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/chartrender.py script allows the agent to write and execute custom Python scripts (B scripts) for rendering charts. This is a core feature for visualization but involves runtime code execution via a subprocess. The skill mitigates this by scanning for forbidden tokens such as socket, eval, and subprocess within the generated code and scrubbing the environment variables to prevent credential leakage.- [COMMAND_EXECUTION]: Multiple scripts within the skill utilize subprocess.run to perform their tasks. scripts/chartrender.py executes custom chart scripts, while smoke_dr.py executes internal skill scripts like evidence.py and aminer_open.py for automated testing and verification.- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted data from the web and the AMiner API.
Ingestion points: Data is ingested via scripts/aminer_open.py and the host's WebFetch tool.
Boundary markers: The instructions define a strict research loop requiring verbatim excerpts, source tagging, and citation-faithfulness verification to prevent the model from obeying instructions hidden in data.
Capability inventory: The skill can execute commands via subprocess.run, perform network calls, and write files to the system.
Sanitization: A token blacklist and environment scrubbing are applied to dynamically generated code.- [EXTERNAL_DOWNLOADS]: The client script scripts/aminer_open.py communicates with the official AMiner API at datacenter.aminer.cn. As this is the vendor's own infrastructure and necessary for the skill's stated purpose, it is documented here as a primary network ingestion point.

deep-research