The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to run tests via the run_test_subprocess activity. This is the primary mechanism used to reproduce and analyze test flakiness. The specific command to be run is provided by the user via input arguments or inferred from the project's test configuration.
[INDIRECT_PROMPT_INJECTION]: The skill processes data including test source code, fixture definitions, and test execution logs (stdout/stderr) to generate hypotheses and synthesize reports.
Ingestion points: Test files, fixtures, and the output captured from test runs are read and processed by subagents.
Boundary markers: The skill relies on structured JSON output schemas for its subagents rather than explicit content delimiters for ingested log data.
Capability inventory: The skill possesses capabilities for shell command execution (run_test_subprocess) and writing artifacts to the local file system (write_artifact).
Sanitization: Log content is interpolated into prompt templates without explicit sanitization, though the use of constrained JSON output for synthesis helps mitigate injection risks.
[EXTERNAL_DOWNLOADS]: The skill's reference documentation provides templates for installing development tools, such as the Go stress tool from its official repository at golang.org. These are documented as standard procedures for diagnosing specific language environments.

flaky-test-diagnoser