The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize external data from various sources such as test results, logs, and exports in formats like CSV, Excel, and JSON. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the processed data. However, the core prompt in prompts/test-reporting.md and the SKILL.md workflow include strong defensive instructions requiring the agent to audit the credibility of supplied materials, separate confirmed facts from inferences, and report missing evidence explicitly, which mitigates the risk of the agent being misled.
Ingestion points: The scripts/parse_formats.py utility and its associated wrappers ingest various file formats into the agent's context.
Boundary markers: The prompt mandates a highly structured output format (Executive Summary, Scope, Risks, etc.) to keep data and analysis separated.
Capability inventory: The skill provides tools for reading and converting files; it does not include tools for automatic network operations or unauthorized file writing.
Sanitization: The instructions mandate that the agent performs manual validation of the evidence provided in the input data.
[COMMAND_EXECUTION]: Several helper scripts in the scripts/ directory, including batch_convert_templates.py and convert_to_csv.py, utilize subprocess.call to execute other local Python scripts. These operations are used for legitimate format conversion tasks, such as turning Markdown reports into CSV or JSON, and are invoked using the local Python interpreter (sys.executable).

test-reporting