The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including requirement documents, test cases, and multi-role reports, creating a surface for indirect prompt injection.
Ingestion points: scripts/run_review.py reads user-provided files in Markdown, JSON, HTML, Word (.docx), and Excel (.xlsx) formats.
Boundary markers: The skill uses NOT_ASSESSED markers in its evidence preparation scripts (scripts/evidence_scaffold.py) and explicitly instructs the agent to "Separate confirmed facts from current assumptions" and "Do not invent requirements or defects the user did not provide" in prompts/test-case-reviewer-plus.md.
Capability inventory: The skill possesses local file read and write capabilities through its helper scripts and can execute shell commands via the agent environment.
Sanitization: scripts/common_formatter.py utilizes xml.sax.saxutils.escape for XML/Word outputs and standard Markdown code fencing to delimit parsed content.
[PRIVILEGE_ESCALATION]: The README.md file suggests using -ExecutionPolicy Bypass when running the PowerShell installation script. While a common practice for local development tools, this bypasses a standard security control for script execution.
[COMMAND_EXECUTION]: The skill provides various Python scripts (scripts/run_review.py, scripts/parse_excel.py, etc.) meant to be executed by the agent to prepare data for analysis. These scripts perform local file operations and data parsing.

test-case-reviewer-plus