The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, specifically test coverage reports (LCOV, JSON, XML) and requirement documents. This creates a potential attack surface for indirect prompt injection where embedded instructions could attempt to subvert the agent's behavior.\n
Ingestion points: Data enters via the parse_coverage_report method in coverage_analyzer.py and the generate_from_requirements method in test_generator.py.\n
Boundary markers: There are no explicit code-level delimiters or warnings instructing the agent to ignore instructions within the processed data.\n
Capability inventory: The skill's capabilities are limited to static analysis, metric calculation, and code stub generation. It lacks dangerous capabilities such as arbitrary shell execution, network exfiltration of local files, or persistent system modifications.\n
Sanitization: While data is parsed into structured objects, the skill does not implement specific filtering or sanitization to remove potential prompt injection payloads.\n- [EXTERNAL_DOWNLOADS]: The skill's documentation references external resources for installation and development workflows.\n
Evidence: The README.md suggests using git clone from GitHub and pip install for development dependencies. These actions are intended for the user to perform and target established, well-known platforms.

tdd-guide