The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through the perf_budget_checker.py utility which generates Markdown reports.
Ingestion points: In scripts/perf_budget_checker.py, the _load_json function reads performance results, scenario names, and descriptions from an external JSON file.
Boundary markers: The resulting Markdown report produced by cmd_report does not include boundary markers or explicit instructions to ignore potentially malicious instructions embedded within the interpolated data.
Capability inventory: The script is capable of reading data files and writing the final Markdown report to a user-specified path on the file system via the pathlib module.
Sanitization: There is no evidence of sanitization, escaping, or schema validation applied to string fields such as service_name or test scenario descriptions before they are interpolated into the report document.

qa-testing-performance