# Next move Use this skill when you arrive at the agent service and need one safe, useful next step before any account or provider action. ## Gather facts privately Collect only the facts the agent is allowed to share. The assessment contract has 27 input collections, including resources, capabilities, provider access, connections, constraints, goals, relationships, nodes, demand, offers, location, network facts, workload, deadlines, budgets, schedules, data policy, approved routes, tools, skills, workflows, repeated work, assets, readiness, rights, authority and current route. Missing or undisclosed facts stay unknown. Never put passwords, tokens, API keys or connector credentials in an assessment. Credentials use a separate authorized connection flow. ## Follow the loop The service follows Discover, Assess, Recommend, Authorize, Execute, Verify, Update and Return. Ask for the smallest useful assessment, then call the shared journey contract. See `docs/v2/GOLDEN-JOURNEYS.md` for the fourteen starting positions and their starting value. When the service is live, the initial tools are: - `assess_agent` - `get_best_path` - `find_free_inference` - `find_optimization` - `check_continuity` - `find_opportunities` Money, account creation, provider enrollment and principal permission are sensitive actions. Stop and request the required authorization before them. ## Communicate results accurately Label evidence as estimated, observed, verified or realized. Say what is unknown and give a recovery route. Never claim guaranteed results or passive income. A researched listing is not an executable integration, and an authorized execution is not verified benefit until its outcome is evidenced. The service is not deployed and this kit is not yet installable from outside; ask the platform owner for the current delivery path.