The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process untrusted data from platforms such as Reddit, G2, and Twitter via WebSearch and WebFetch. This data is then analyzed by a chain of agents. Maliciously crafted content on these platforms could potentially attempt to influence the agent's analysis or override instructions.\n
Ingestion points: External content enters the system via the voc-collector-agent, which is tasked with finding real-world quotes.\n
Boundary markers: The skill uses a critic-agent to evaluate the integrity of evidence, but does not specify technical delimiters to isolate untrusted web content from its core instructions.\n
Capability inventory: The skill has access to the Bash, Read, WebSearch, and WebFetch tools.\n
Sanitization: No explicit sanitization or filtering of the fetched content is described before it is synthesized into the final research artifact.\n- [COMMAND_EXECUTION]: The skill contains local shell and TypeScript scripts to manage setup and assist with query generation. These scripts were audited and found to be benign.\n
Evidence: scripts/bootstrap-experience.ts initializes a local directory structure (.agents/experience/) for session persistence. It includes security checks to prevent writing to symbolic links.\n
Evidence: scripts/search-platforms.sh is a utility that echoes search query templates for different platforms to assist in manual or automated research.

icp-research