The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to interact with the local filesystem and external CLI tools (git, gh, paseo). It instructs the agent to perform potentially destructive operations including git worktree remove and paseo workspace archive. The instructions mitigate this by mandating preflight checks for uncommitted changes, merged statuses, and active agent sessions, as well as requiring a single batch confirmation from the user before any changes are applied.- [INDIRECT_PROMPT_INJECTION]: The skill instructions define a process for fetching external metadata that is then used to construct shell commands, creating a vulnerability surface.
Ingestion points: The agent fetches issue and pull request titles from the GitHub CLI tool (gh), which retrieves data from a remote, attacker-controlled environment.
Boundary markers: The skill does not provide specific instructions to use delimiters or ignore markers when interpolating these external titles into agent prompts or command strings.
Capability inventory: The agent has access to the Bash tool, allowing it to perform file system writes, register new workspaces, and rename existing ones based on the ingested content.
Sanitization: There are no explicit requirements in the SKILL.md for the agent to sanitize, filter, or escape the titles fetched from GitHub before they are passed as arguments to shell commands such as paseo workspace rename.

paseokit