The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute attach and gh CLI commands for media management and uses curl for API interactions. It employs defensive patterns like using temporary header files (-H @"$hdr") to ensure tokens are not exposed in command-line arguments.
[EXTERNAL_DOWNLOADS]: The skill identifies and references several vendor-owned installation paths, including the @uinaf/attach-cli Node.js package, a Homebrew tap, and a GitHub CLI extension.
[INDIRECT_PROMPT_INJECTION]: The skill presents a potential surface for indirect prompt injection through the following components:
Ingestion points: Untrusted data enters the context via the attach put command which reads local files, and through HTTP responses from the enrollment and upload endpoints (e.g., attach.uinaf.dev).
Boundary markers: The skill encourages the use of structured output (--json) to minimize ambiguity, although specific prompt delimiters for file content are not defined.
Capability inventory: The skill allows for shell command execution, network operations via curl and the CLI, and file system writes for temporary header management.
Sanitization: Includes strict URL validation for deletion targets, the use of --dry-run for local validation, and verification of resource removal using curl -sI.

attach-cli