The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill employs npx to download and execute packages from the NPM registry without user confirmation (-y / --yes flags).
The SKILL.md instructs the agent to run npx -y @nomadamas/k-skill@0 instruct bunjang-search as the mandatory first step to acquire the "full instructions," effectively moving the primary logic to a remote source.
The SKILL.md also suggests running npx -y @nomadamas/k-skill@0 update to keep the environment current, which involves executing remote update scripts.
The instruction.md file defines its core policy around executing npx --yes bunjang-cli, a third-party tool maintained by an external GitHub user (pinion05).
[COMMAND_EXECUTION]: The skill is heavily reliant on shell command execution to perform all its core functions, including authentication, searching, and data collection.
[EXTERNAL_DOWNLOADS]: At runtime, the skill triggers multiple downloads from the NPM registry to fulfill dependencies for @nomadamas/k-skill and bunjang-cli.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests and processes untrusted data from the Bunjang marketplace.
Ingestion points: Marketplace item titles, descriptions, seller reviews, and status fields are retrieved via bunjang-cli and presented to the agent (see instruction.md Search/Detail flows).
Boundary markers: The instructions do not mention the use of delimiters or specific warnings to the agent to ignore embedded instructions within marketplace content.
Capability inventory: The skill possesses the ability to execute shell commands (npx), write files to the filesystem (e.g., --output artifacts/), and perform network operations via the CLI utility.
Sanitization: No evidence of sanitization, escaping, or strict schema validation for the external marketplace content is provided in the skill's instructions.

bunjang-search