The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes bundled Python scripts (sync.py, test_sync.py) to manage agent synchronization and verification. It also identifies and suggests running repository-specific tools like npm audit, gitleaks, and cargo test as part of its repository discovery and quality gate verification processes. All commands are targeted at local repository management and standard developer tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes a wide array of untrusted repository data, including build manifests, CI configurations, and project documentation (e.g., CLAUDE.md). This creates an indirect prompt injection surface. The skill mitigates this by providing detailed instructions in roles.yaml that require subagents to verify claims against the actual code and utilize specific boundary markers for context handoff, reducing the likelihood of the agent blindly following instructions embedded in repository data.
[DYNAMIC_EXECUTION]: The skill dynamically generates subagent definition files (.claude/agents/*.md) and workflow manifests (.claude/agent-team.md) based on the repository's discovered state and a versioned role library. These files are used to configure the subagent environment for subsequent spawns. The generation process uses a strictly defined template and a local YAML library, maintaining control over the instruction set.

agent-team