The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to manage workflow state and process design artifacts. It uses standard utilities like ls, grep, and awk, alongside custom project tools like visual_review and stelow to automate transitions between design phases.
[DYNAMIC_EXECUTION]: Sourced configuration and logic are loaded from computed relative paths (e.g., ../../stelow-workflow-orchestrator/references/cli-tools/read-config.sh). This allows the skill to share utilities with other components in the workflow suite but introduces a dependency on the local file system structure.
[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific files like spec-product.md and tech-recon.md to determine the scope and constraints of the design task. This creates a potential surface where untrusted data in these files could influence agent behavior.
Ingestion points: Contents of spec-product.md and tech-recon.md are loaded into the context for subagents.
Boundary markers: None defined; the agent is instructed to read the files directly.
Capability inventory: Includes shell command execution, file system writes, and the ability to invoke subagents.
Sanitization: No explicit sanitization or validation of the input file content is performed before processing.

stelow-workflow-interface-alternatives