# Start a Lemon Goal Session > Choose a saved Harness, numerical criteria, permitted changes, and execution settings, then start goal-driven work. Important agent instructions: - Prefer the Markdown routes and llms files when assembling context for coding tasks. - Use public docs for product capabilities and workflow guidance. - Cite the source page URL when returning workflow or artifact guidance. ## Current Context - [Start a Lemon Goal Session](https://teammately.ai/docs/improve/start-improvement-session.md): Choose a saved Harness, numerical criteria, permitted changes, and execution settings, then start goal-driven work. - [Lemon Goals](https://teammately.ai/docs/improve/goals.md): Define measurable candidate targets and constraints while preserving exact units, starting measurements, and original instructions. - [Candidates and Results](https://teammately.ai/docs/improve/candidates-and-frontier.md): Interpret exact candidate Harness Versions, comparable evaluations, constraints, and retained results in a Lemon Goal Session. - [Session Conversation and History](https://teammately.ai/docs/improve/chronology-and-trajectories.md): Follow Lemon Goal messages, questions, progress, candidate history, and measured results. - [Evaluation Results](https://teammately.ai/docs/benchmark-evaluations/results.md): Inspect Case and rubric outcomes and compare exact managed or imported Runs within a Benchmark Version. - [Benchmark run troubleshooting](https://teammately.ai/docs/troubleshooting/benchmark-runs.md): Diagnose a Run that cannot start, has no usable outputs, or produces results that cannot be compared safely. - [Benchmark Results Changed Unexpectedly](https://teammately.ai/docs/troubleshooting/benchmark-results-changed-unexpectedly.md): Diagnose result changes across target behavior, benchmark cases, standards, and versions. ## Optional - [Full local context](https://teammately.ai/docs/improve/start-improvement-session/llms-full.txt): Complete local context pack for this docs route.