Authentication and enablement
- No external integration. It reads Trails in the running agent’s workspace.
- Off by default. Not seeded into new agents.
- An agent sees only its own workspace’s Trails, and only approved rules.
Inputs
query(required): what you want to know, in plain words.trails(optional): Trail names or ids. Leave it out to search every Trail in the workspace. One name searches that Trail. Several names search those Trails.domains(optional): domain names. Only rules in a domain with that name are searched. The same name in two Trails matches both. Names come fromlist_trail_domains_and_tags.clusters(optional): cluster names, matched the same way. Combines withdomains: a rule must be in both.domainsandclustersnarrow what is searched. They do not change how the remaining rules are ranked. Rules that are not in a domain are left out when either filter is set.
Output
One ranked list. Each rule includes its text, rule class, rank, the Trail it came from (name and id), and what in the query it matched. When two returned rules conflict, both are shown with their Trails and the conflict is named. The tool does not pick a winner. An empty result means nothing stored matched the query. A Trail, domain, or cluster name that does not exist returns the valid names instead of a result.Limits and side effects
- Read-only. It does not create, edit, or approve rules.
- The search is recorded in the Activity tab of each Trail it returned rules from.
- Searching one Trail returns the same rules as
get_brain_contexton that Trail. - Conflicts between rules in different Trails are not checked when a rule is added. They show up only when a search returns both.
Expected errors
- Missing
query. - An unknown Trail, domain, or cluster name. The error lists the valid names.
- Workspace missing from the call.