The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands for file management (cp, rm, mkdir), Git operations (git worktree), and process control (kill). It also runs internal CLI scripts using the Node.js runtime.
[COMMAND_EXECUTION]: The skill executes a PowerShell script (prevent-sleep.ps1) using the -ExecutionPolicy Bypass flag on Windows to override security restrictions, which is a high-risk command execution pattern.
[REMOTE_CODE_EXECUTION]: The skill performs dynamic loading of Node.js modules in scripts/lib/arch-snapshot.mjs using import() with computed paths. It specifically targets paths outside the skill's own directory (../../../mcp/hex-graph-mcp/lib/store.mjs) to load external libraries.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted data from kanban_board.md and task tracking APIs (Linear and GitHub). This data influences orchestration logic and is passed to delegated sub-skills.
Ingestion points: Story titles, descriptions, and Epic metadata from kanban_board.md and API responses.
Boundary markers: Lacks explicit delimiters or instructions to ignore embedded commands within ingested story content.
Capability inventory: Significant subprocess (shell, powershell, git) and file system capabilities are used across all runtime scripts.
Sanitization: Content from story descriptions is not sanitized or escaped before being processed by the orchestrator.

ln-1000-pipeline-orchestrator