The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git commands for worktree management and triggers external AI agent CLIs (Claude, Codex, etc.) to perform coding tasks. Standard shell utilities like curl and jq are used for status monitoring.
[EXTERNAL_DOWNLOADS]: The skill relies on npx to fetch and run the vibe-kanban tool from the NPM registry. It also references official Docker images for containerized deployment.
[INDIRECT_PROMPT_INJECTION]: The skill passes user-supplied task descriptions to AI agent CLIs, presenting a potential surface for indirect prompt injection.
Ingestion points: Task descriptions provided by the user in the workspace creation UI (documented in SKILL.md).
Boundary markers: None identified; input is passed directly to agents.
Capability inventory: The skill has broad command execution and file modification capabilities via git worktrees (scripts/cleanup.sh, scripts/start.sh).
Sanitization: No input sanitization logic is present in the provided shell automation scripts.
[PRIVILEGE_ESCALATION]: The docker-compose.yml template suggests mounting /var/run/docker.sock, which provides the container with high-level access to the host's Docker engine.
[PERSISTENCE]: The mcp-setup.sh script automates the addition of the Vibe Kanban server to agent configuration files (e.g., ~/.claude/claude_desktop_config.json), ensuring the tool is available across agent sessions.

vibe-kanban