The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell command snippets in its workflow definition (e.g., ls, cat, grep, head). These commands are used to inspect the file system and track project tasks, establishing a pattern of direct command execution based on project-specific paths and file contents.
[PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection as it ingests untrusted data from project specification files and passes it to various sub-agents without safety boundaries. * Ingestion points: Project specification files at project-specs/[project]-setup.md and task lists at project-tasks/*-tasklist.md. * Boundary markers: Absent. The skill passes content from these files directly into the instructions for spawned agents without using delimiters or instructions to ignore embedded commands. * Capability inventory: The orchestrator has the authority to spawn high-privilege sub-agents, including DevOps Automator, Backend Architect, and engineering-senior-developer, which can perform broad system and infrastructure operations. * Sanitization: None. There is no evidence of input validation, escaping, or filtering for the data read from external project files.

agents-orchestrator