The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from project specification files which may contain instructions targeting the orchestrator or its sub-agents.
Ingestion points: Reads content from files located at project-specs/[project]-setup.md and project-tasks/[project]-tasklist.md throughout the workflow phases.
Boundary markers: Absent. The skill does not implement delimiters or instructions for the agents to ignore potential prompt injections embedded within the project files.
Capability inventory: The orchestrator has the capability to spawn numerous specialized agents (e.g., Frontend Developer, Backend Architect, EvidenceQA) and execute shell commands (ls, cat, grep).
Sanitization: Absent. There is no evidence of validation or sanitization of the input data before it is passed to agents or used in shell commands.
[COMMAND_EXECUTION]: The skill uses local shell commands to manage file state and track progress during the autonomous workflow.
Evidence: The workflow includes commands such as ls -la, cat, and grep to verify file existence and count tasks (e.g., TASK_COUNT=$(grep -c "^### \[ \]" ...)). These are used for project management within the local environment.

agency-agents-orchestrator