The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions specify the direct shell execution of bun i and bun run build. These commands automatically run lifecycle and build scripts defined in the repository's package.json.
[REMOTE_CODE_EXECUTION]: The skill facilitates the checkout of a user-controlled Git reference (ref) followed by code execution via bun. This pattern is vulnerable to arbitrary code execution if an attacker provides a reference to a malicious branch that contains harmful commands in the postinstall, preinstall, or build script fields.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via untrusted repository references.
Ingestion points: User-provided Git reference ('ref') in SKILL.md used for checkout operations.
Boundary markers: None present to limit the scope of the checkout or validate the reference.
Capability inventory: Shell command execution capability via bun scripts as defined in SKILL.md.
Sanitization: No validation or sanitization is performed on the user-provided reference prior to its use in shell commands.

checkout