The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes external content from arXiv, PubMed, and policy databases, which can act as a vector for indirect prompt injection if those sources contain malicious instructions.
Ingestion points: Paper metadata and full-document Markdown retrieved from external corpora in SKILL.md.
Boundary markers: The skill does not provide specific instructions to the agent to use delimiters or sanitization when interpreting the retrieved content.
Capability inventory: The agent has access to shell command execution through the npx runner.
Sanitization: The documentation recommends using argument arrays with spawnSync to mitigate the risk of shell command injection from user queries.
[COMMAND_EXECUTION]: The skill operates by executing CLI commands through the ii-commons tool to perform retrieval tasks.
[REMOTE_CODE_EXECUTION]: The skill uses npx to fetch and run the @intelligentinternet/ii-commons package from the npm registry. This is the intended method for providing the skill's functionality.
[EXTERNAL_DOWNLOADS]: The skill connects to the vendor's API at commons.ii.inc to search and retrieve research metadata and documents.

ii-commons