The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data including scope.json, campaign indices, and progress ledgers to generate reports.
Ingestion points: Reads campaign data files including scope.json, progress ledgers, and stage packages from the local directory.
Boundary markers: No explicit data delimiters are used, but the instructions require strict adherence to standard definitions and project-specific guidance files (../../docs/regression-definition.md).
Capability inventory: The skill executes a local script report.py and writes output artifacts to the assigned reports directory.
Sanitization: The instructions explicitly forbid the agent from repairing decisions, inventing evidence, or editing hashes to pass validation, which serves as a prompt-level constraint against untrusted data influence.
[COMMAND_EXECUTION]: The skill invokes a local Python script using the uv tool.
Evidence: uv run --no-project ../../scripts/report.py
The script is located within the repository's internal file structure and is used for its intended purpose of rendering reports in HTML, Markdown, and JSON formats.

report-regressions