The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/onlform_jimureport.py contains a hardcoded signing key SIGNATURE_SECRET ('dd05f1c54d63749eda95f9fa6d49v442a'). This key is used to generate an MD5-based signature for JimuReport API endpoints to authenticate requests.
[DYNAMIC_EXECUTION]: The utility script scripts/onlform_defval.py uses the Python eval() function to resolve custom default value expressions (e.g., {{Math.random()}}) found in form configuration files. While the script attempts to isolate the environment by providing a restricted safe_globals dictionary and removing __builtins__, the execution of arbitrary strings as code is inherently risky.
[INDIRECT_PROMPT_INJECTION]: The skill routinely ingests metadata and record data from remote JeecgBoot API endpoints, which creates a surface for indirect prompt injection if those endpoints are compromised or malicious.
Ingestion points: Metadata retrieval in scripts/onlform_creator.py (querying field lists) and record retrieval in scripts/onlform_data.py (querying data for display or export).
Boundary markers: Absent; data retrieved from external APIs is printed into the agent's context as raw JSON or text without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill allows for network operations via urllib.request, file system writes to create temporary JSON configurations, and the execution of Python scripts via the agent's shell tools.
Sanitization: There is no evidence of sanitization or strict schema validation performed on the API responses before the content is presented back to the AI agent.

jeecg-onlform