The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/codegen.py uses the Python subprocess module to execute the Java compiler (javac) and the Java runtime (java). This is used to build and run the internal template engine (FtlRunner.java) included with the skill.
[CREDENTIALS_UNSAFE]: The SKILL.md file contains several command-line examples for interacting with MySQL that include hardcoded default developer credentials (-uroot -proot).
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection where data processed by the skill could influence agent behavior.
- Ingestion points: User-provided natural language requirements and database schema information (DDL) retrieved via shell commands in SKILL.md enter the agent's context.
- Boundary markers: Absent. The skill does not provide specific instructions to the model to ignore or delimit embedded instructions within the ingested data.
- Capability inventory: The skill can write arbitrary files to user-specified directories (--backend-root, --frontend-root) and execute local shell commands via codegen.py and FtlRunner.java.
- Sanitization: The normalize_ctx logic in codegen.py provides structure and defaults but does not sanitize input for potential natural language injection patterns.

jeecg-codegen-new