The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a set of scripts and templates for the agent to manage the development lifecycle, including git operations (worktrees, merging, status), container management (Docker), and deployment (Kubernetes/kubectl). These commands are standard for a technical orchestration skill and are used within the scope of intended development tasks.
[EXTERNAL_DOWNLOADS]: The workflow relies on standard package managers like npm and pip to install dependencies and run security audits. These operations target well-known, trusted package registries and are a routine part of modern software development.
[INDIRECT_PROMPT_INJECTION]: The skill's 'Observer' agents are designed to monitor and analyze code and contracts contributed to the project. This establishes a surface where the agent processes potentially untrusted data that could contain adversarial instructions. However, this risk is inherent to any tool designed for code analysis and is mitigated by the skill's focus on structured reporting.
Ingestion points: The skill reads from '.claude/CONTRACT.md', builder branches, and session context files like 'PROJECT_CONTEXT.md'.
Boundary markers: The instructions do not specify explicit delimiters or isolation markers when the observer agents process the code content.
Capability inventory: The agents have the ability to execute shell commands, write to the filesystem, and spawn additional specialized agents.
Sanitization: There is no mention of specific sanitization or filtering of instructions that might be embedded in the code being observed.

workflow-orchestrator-skill