The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs automated scanning of sensitive shell configuration files, including ~/.bashrc, ~/.bash_profile, ~/.zshrc, and ~/.profile. It uses regular expressions to extract specific environment variables related to API credentials (ARK_* and VOLCSTACK_*). While the skill targets specific keys and implements logging redaction, accessing user shell profiles is a high-privilege read operation.
[COMMAND_EXECUTION]: The script scripts/auth_resolver.py utilizes subprocess.run to execute pip install for the volcengine-python-sdk package. Executing shell commands to manage system or environment packages at runtime is a sensitive operation.
[EXTERNAL_DOWNLOADS]: The skill dynamically downloads the volcengine-python-sdk from the official Python package registry if it is not detected in the environment. This is an official dependency from the vendor ByteDance.
[DYNAMIC_EXECUTION]: The skill uses importlib.import_module to dynamically load SDK components after runtime installation. It also uses dynamic attribute access (getattr) to dispatch calls to different provider implementations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content from social media platforms and AI-generated business leads, which could contain adversarial instructions.\n
Ingestion points: Data enters the agent context via pull_post and query_clue_info functions in scripts/client.py, which retrieve content from the Volcengine Insight API.\n
Boundary markers: None identified; external data is interpolated into the agent context without explicit delimiters or instructions to ignore embedded commands.\n
Capability inventory: The skill possesses the capability to perform network operations, write to the persist/ directory, execute shell commands via pip, and perform dynamic imports.\n
Sanitization: The skill attempts to parse retrieved clue text as JSON but does not perform content-based filtering or sanitization of the resulting data.

byted-market-insight-agent