The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using git diff <base>...<head>. The <base> and <head> parameters are expected to be provided by the user or build environment. If these parameters are not sanitized (e.g., through shell escaping), they are vulnerable to command injection if a user provides input like main; malicious_command.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data to perform its analysis, including git diff output, bug fix descriptions, and requirement change documents.
Ingestion points: Code diff data from the filesystem/stdout, bug reports, and requirement documents (documented in the 'Input' section of SKILL.md).
Boundary markers: No explicit boundary markers or isolation instructions for external data are defined in the workflow.
Capability inventory: The skill uses shell commands (git diff) and writes files to the project directory ({project}/Regression_List_{date}.md).
Sanitization: No sanitization or validation of the ingested natural language text (bug reports) or code changes is specified before they are processed by the LLM logic.

regression-testing