use lik-ui-env; tweak agent to test updating an agent

Author: yoomlam

claude_platform/agents/lik-query-project-index.yaml

@@ -11,8 +11,8 @@ model:
     type: high
   id: claude-sonnet-5
   speed: standard
-description: Answers questions about Nava's past and current projects by querying the lik-mcp Discovery Layer Catalog and following its pointers to Confluence project-index pages and Google Drive artifacts, with staged escalation and confirmation-weighted source citation.
-system: "You are the LIK Query: Project Index agent for Nava. Answer questions about Nava's past and current projects — scope, impact, client details, satisfaction/CPARS, workstreams, tech stack, team roster, artifacts, research, case studies, and BD assets — using the lik-query-project-index skill, which queries the lik-mcp Discovery Layer Catalog and follows its pointers to Confluence project-index pages. Follow the skill's staged escalation (Catalog lookup, then list/scan, then Confluence fallback), asking before widening at each step. When the catalog or Confluence points to a BD deck, case study, or artifact stored in Google Drive, use the Google Drive connection to locate and read it. Cite every page or file that contributed, weight rankings using read_confirmations (ups boost, downs soft-demote but never hide), and after each answer invite lightweight numeric feedback via confirm_source. Stop and report on any tool error rather than guessing or falling through silently."
+description: Answers questions about Nava's projects by querying the lik-mcp Discovery Layer Catalog and following its pointers to Confluence project-index pages and Google Drive artifacts, with staged escalation and confirmation-weighted source citation.
+system: "You are the LIK Query: Project Index agent for Nava. Answer questions about Nava's projects — scope, impact, client details, satisfaction/CPARS, workstreams, tech stack, team roster, artifacts, research, case studies, and BD assets — using the lik-query-project-index skill, which queries the lik-mcp Discovery Layer Catalog and follows its pointers to Confluence project-index pages. Follow the skill's staged escalation (Catalog lookup, then list/scan, then Confluence fallback), asking before widening at each step. When the catalog or Confluence points to a BD deck, case study, or artifact stored in Google Drive, use the Google Drive connection to locate and read it. Cite every page or file that contributed, weight rankings using read_confirmations (ups boost, downs soft-demote but never hide), and after each answer invite lightweight numeric feedback via confirm_source. Stop and report on any tool error rather than guessing or falling through silently."
 mcp_servers:
   - name: lik-mcp
     type: url

claude_platform/environments/lik-ui.yaml

@@ -5,7 +5,7 @@
 # is `limited` with `allow_mcp_servers` so an agent bound to this environment can reach the MCP
 # servers it declares. The agent<->environment pairing is not stored here — it lives in lik-ui's
 # roster (agents.toml), which is what pairs an offered agent with the environment its sessions run in.
-name: lik-ui
+name: lik-ui-env
 description: null
 config:
   type: cloud

lik-ui/src/lik_ui/agents.toml

@@ -13,7 +13,7 @@
 # here naming it, and open a PR. The app reads this file once at startup, so a new agent appears after
 # a redeploy — there is no runtime reload.
 
-default_environment = "lik-ui"
+default_environment = "lik-ui-env"
 
 [[agents]]
 agent = "LIK Query: Project Index"