The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute git diff and git status to analyze code changes. It also utilizes the mcp-gitlab-usage tool to post review comments and verdicts to GitLab issues when in Issue Review mode. These operations are restricted to the skill's primary purpose of code auditing and reporting.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data in the form of code diffs and GitLab issue context.
Ingestion points: Data enters the agent context through git diff outputs and issue synthesis provided by callers like ns-execution-gitlab-issue (SKILL.md).
Boundary markers: The instructions include explicit exclusions such as "Do NOT parse requirements.md" and "Focus modified files + surrounding context" to limit the processing scope (SKILL.md).
Capability inventory: The skill possesses read-only access to the filesystem (for git commands) and network-writing capability via the GitLab MCP tool for posting comments (SKILL.md).
Sanitization: The skill relies on the agent's internal safety protocols and the defined read-only constraints to prevent execution of embedded instructions.
[EXTERNAL_DOWNLOADS]: The skill depends on ns-harness and references sibling components like ns-judge and ns-langgraph-agents using relative paths (e.g., ../../ns-harness/references/session-boot.md). These references point to internal project resources within the vendor's ecosystem rather than unknown external repositories.

ns-reviewer