The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from project files to drive the execution loop and subagent prompts.
Ingestion points: The orchestrator reads {product_root}/docs/versions/{version_san}/version-roadmap.md and {product_root}/docs/versions/{version_san}/requirements.md to determine the execution order and task content.
Boundary markers: The instructions do not define clear delimiters or escaping mechanisms when interpolating data from these files into the subagent dispatch prompts.
Capability inventory: The skill has the authority to perform git operations (commits, branch checkouts), dispatch subagents with implementation permissions, and execute unit/integration tests.
Sanitization: There is no evidence of validation or sanitization for the content of the roadmap or requirements files before they are used to influence agent behavior.
[COMMAND_EXECUTION]: The skill performs automated command execution through git operations and testing frameworks.
Git Operations: The skill manages branches via mcp-gitlab-usage and performs automated Conventional Commits after each slice.
Test Execution: The orchestrator is instructed to run unit and integration tests to validate subagent results, which involves executing code within the project environment.

ns-execution-orchestrator