The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a workflow where coverage gap data is parsed from project files and passed to a 'test-generator' agent. This creates a risk of indirect prompt injection if project artifacts are compromised.
Ingestion points: coverage.json, coverage.info, .qt-test.json, and project configuration files (pyproject.toml, setup.cfg).
Boundary markers: No explicit delimiters are provided in the suggested handoff prompt to separate coverage data from instructions.
Capability inventory: The downstream 'test-generator' agent is designed to generate source code. The skill's own scripts (run-coverage.sh) execute system build and test tools.
Sanitization: The extraction logic in run-coverage.sh and references/*.md uses basic regex and JSON parsing without sanitization of the content before it is included in prompts.
[COMMAND_EXECUTION]: The skill provides a portable shell script (templates/run-coverage.sh) and a GitHub Actions template (templates/qt-coverage.yml) that execute potentially dangerous system commands including cmake, ctest, lcov, and pytest to perform build and instrumentation tasks.
[EXTERNAL_DOWNLOADS]: Documentation and templates provide instructions for installing external dependencies from public registries using apt-get, dnf, brew, and pip.

qt-coverage-workflow