The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local research harness using bun run src/cli.ts found in the runtime/wide-search directory of the repository.
[EXTERNAL_DOWNLOADS]: Uses built-in WebSearch and FetchURL tools to gather research data from various external websites.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface due to its processing of untrusted web content.
Ingestion points: Data is ingested via the WebSearch and FetchURL tools (as described in references/wide-search-mode.md).
Boundary markers: Subagents use a prompt_template with {{item}} delimiters, but lack explicit instructions to disregard instructions within the ingested data.
Capability inventory: The skill uses high-privilege tools including AgentSwarm, WebSearch, FetchURL, Read, Write, and Bash (documented in SKILL.md).
Sanitization: Employs source quality rules (Relevance, Authority, Freshness) and verification checks (identifying unsupported or conflicting claims) to help sanitize research results.

kimi-agent-swarm-cli