The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and utilize .wtp.yml configuration files located within repositories. If an agent processes a malicious repository, these files could contain shell hooks or instructions designed to influence the agent's behavior or execute unwanted actions during the worktree setup process.
Ingestion points: The agent reads .wtp.yml from the repository root (SKILL.md).
Boundary markers: No specific delimiters or safety instructions are provided to the agent to treat the contents of .wtp.yml as untrusted.
Capability inventory: The skill uses wtp exec and shell commands via post_create hooks, allowing for arbitrary command execution.
Sanitization: There is no evidence of sanitization or validation of the commands defined in the configuration files.
[DYNAMIC_EXECUTION]: The skill promotes the use of lifecycle hooks (post_create) within the .wtp.yml file, which allows for the automatic execution of shell commands when a new worktree is created. It also provides patterns for shell integration using eval to initialize the wtp environment.
[COMMAND_EXECUTION]: The skill relies on the execution of the wtp CLI and standard Git commands to manage worktrees. It provides the agent with the ability to run arbitrary user-specified or config-specified commands inside a worktree via wtp exec <name> -- <command>.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the wtp binary from external sources, including a Homebrew tap (satococoa/tap/wtp) and a Go package registry targeting a GitHub repository (github.com/satococoa/wtp/v2/cmd/wtp@latest).

jb-worktree