The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection and data processing risks due to its handling of untrusted external input. * Ingestion points: Untrusted data enters the skill via coverage reports (XML, JSON, LCOV) in scripts/coverage_analyzer.py and feature requirements (JSON) in scripts/test_generator.py. * Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to ignore embedded commands within the processed data. * Capability inventory: The skill is capable of generating executable test files and performing analytical tasks on source code. * Sanitization: There is no evidence of sanitization or character escaping for input data before it is processed or used in code generation. * Evidence: In scripts/coverage_analyzer.py, the _parse_xml method utilizes xml.etree.ElementTree.fromstring to parse coverage reports. This standard library function is vulnerable to XXE attacks, which could be exploited by a malicious report to read local files or conduct SSRF. * Evidence: In scripts/test_generator.py, methods such as _generate_jest_stub perform direct string interpolation of user-provided requirement fields into code templates. The lack of quote escaping allows for potential code injection into resulting test files if the requirements are maliciously crafted.

tdd-guide