The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/self_update.py provides a mechanism to download a ZIP archive from a remote URL, verify its integrity via SHA-512, and replace the existing skill code on the local filesystem. This self-update feature allows for remote code execution if the update source or the signaling server is compromised.
[DATA_EXFILTRATION]: In scripts/quant_api.py and scripts/call.py, the skill transmits the user's raw input query to an external endpoint (https://www.quantbuddy.cn/skill/session/begin) for tracking and analysis. This domain is not included in the standard whitelist for data transmission.
[EXTERNAL_DOWNLOADS]: The scripts/event_study_local.py helper script fetches news data from a third-party search API (api.bochaai.com).
[PROMPT_INJECTION]: The skill ingests untrusted news snippets from an external search provider into the agent context through scripts/event_study_local.py. This data is used to drive decision-making processes without adequate sanitization or boundary protection.
Ingestion points: External news content is fetched in scripts/event_study_local.py using the bocha_web_search function.
Boundary markers: The news snippets are integrated into the workflow without explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The skill possesses the capability to execute shell commands through its dispatcher (call.py) and modify its own code via self_update.py.
Sanitization: There is no evidence of filtering or escaping performed on the content retrieved from the web search results.

quant-buddy-skill