The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill relies on the command npx --yes daiso, which automatically downloads and executes code from the NPM registry at runtime. It also provides a fallback method involving cloning a repository (https://github.com/hmmhmmhm/daiso-mcp.git), installing dependencies, and executing the build output with Node.js.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to fetch external scripts and binaries from GitHub and NPM. These sources (author 'hmmhmmhm') are not associated with a known trusted organization or verified service provider.
[COMMAND_EXECUTION]: The skill requires the execution of multiple shell commands to interact with the retail search tool, including npx, npm install, npm run build, and node. This provides a direct path for executing code fetched from external sources.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external APIs (store and product search results) using the daiso CLI. This creates a surface where malicious instructions could be embedded in the API responses and processed by the agent.
Ingestion points: Data enters the context via the JSON output of npx --yes daiso get commands in SKILL.md.
Boundary markers: None provided in the instructions.
Capability inventory: Subprocess execution via npx, git clone, npm, and node in SKILL.md.
Sanitization: No sanitization or validation of the external API output is specified before the agent summarizes the data.

olive-young-search