The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md require the agent to fetch instructions and files from an external source using npx -y @nomadamas/k-skill@0.
[REMOTE_CODE_EXECUTION]: The skill uses npx to execute code from the @nomadamas/k-skill package, which is an untrusted external source, to determine the instructions the agent should follow.
[COMMAND_EXECUTION]: The instruction.md file directs the agent to perform package installation (npm install gongsijiga-search) and shell command execution using the Node.js CLI.
[DYNAMIC_EXECUTION]: The skill uses node -e to dynamically evaluate and execute JavaScript code that imports and runs functionality from the unverified gongsijiga-search package.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection as it processes and displays data retrieved from an external NPM package without explicit sanitization or boundary markers described in the instructions.
Ingestion points: Data is ingested through the lookupGongsijiga function call in the gongsijiga-search package (instruction.md).
Boundary markers: No boundary markers or 'ignore embedded instructions' warnings are present for the external data.
Capability inventory: The skill uses npx, npm install, and node -e (SKILL.md, instruction.md).
Sanitization: There is no mention of sanitization or filtering for the external content before it is displayed to the user.

gongsijiga-search