The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file scripts/get_data.py contains a hardcoded default API key for the Eastmoney service: em_fjFqd4YB6Cqs52LF48XWbMDdLNq6MyNg. Hardcoding credentials in source code exposes the service account to unauthorized use and exfiltration.
[COMMAND_EXECUTION]: The execution logic defined in SKILL.md (Step 2) instructs the agent to run a shell command using string interpolation: python3 {baseDir}/scripts/get_data.py --query "{{topic}}". This pattern is vulnerable to command injection if the extracted {{topic}} contains shell metacharacters (e.g., ;, &, |, or backticks), potentially allowing an attacker to execute arbitrary code on the host system.
[DATA_EXFILTRATION]: While the script targets a well-known financial service domain (ai-saas.eastmoney.com), the presence of hardcoded credentials facilitates unauthorized data access or service abuse under the author's identity.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8).
Ingestion points: User-provided industry queries are processed to extract keywords (SKILL.md Step 1).
Boundary markers: No delimiters or safety instructions are used to sanitize the extracted {{topic}} before it is passed to the shell.
Capability inventory: The skill's scripts have the capability to perform network POST requests and write files to the local file system (scripts/get_data.py).
Sanitization: There is no validation or escaping performed on the {{topic}} variable before it is interpolated into the shell execution string.

industry-research-report