The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted application source code during its assessment phase (e.g., in references/tech-stack-detection.md and references/blocker-detection.md). This ingestion point is managed through strict process isolation and rules enforcing read-only access to the source tree and environment. Mitigation: The skill implements 'Assessment_Phase — read-only invariants' and an 'Execution_Gate' that prevents untrusted input from triggering destructive operations without explicit human intervention.
[COMMAND_EXECUTION]: The skill utilizes several system tools including terraform, docker, and aws-cli to perform migration tasks (references/deploy-verify-handoff.md). Every action class is gated behind mandatory user confirmation as described in the SKILL.md Execution_Gate workflow. Mitigation: The requirement for per-action-class confirmations and the plan destroy check ensures continuous user oversight.
[EXTERNAL_DOWNLOADS]: The skill references various external documentation and tools such as the .NET Upgrade Assistant and IBM Transformation Advisor (references/rearchitect-path.md). These links point to well-known technology providers and are used to provide the user with established migration paths. Mitigation: No automated execution of external code from these sources occurs; they are presented as informational resources or manual accelerators.

ecs-modernize