The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using npx --yes bunjang-cli. Use of the --yes flag bypasses confirmation prompts, allowing for automatic execution of the third-party tool.
[EXTERNAL_DOWNLOADS]: The skill triggers the download and installation of the bunjang-cli package from the npm registry (npmjs.com) during runtime. This introduces a dependency on a third-party package managed by an external individual.
[REMOTE_CODE_EXECUTION]: By executing a third-party package from a public registry without pinning a specific version or cryptographic hash, the skill is vulnerable to supply chain attacks. If the bunjang-cli package were compromised, it would result in arbitrary code execution on the user's system.
[PROMPT_INJECTION]: The skill processes untrusted content (item descriptions and titles) from a public marketplace. This creates an Indirect Prompt Injection surface where an attacker could place instructions in a product listing to influence the agent's behavior.
Ingestion points: Data enters via the output of bunjang-cli search and bunjang-cli item get commands.
Boundary markers: The instructions do not specify any delimiters or warnings to treat marketplace data as untrusted text.
Capability inventory: The agent has capabilities to write files (--output to artifacts/), execute shell commands (npx), and communicate over the network (via the CLI tool).
Sanitization: No sanitization or validation of the marketplace data is performed before the agent processes or analyzes it.

bunjang-search