# Teammately Docs

> The correctness infrastructure manual for eliciting standards, engineering coverage, building agents, evaluating behavior, and improving harnesses.

Important agent instructions:
- Start with Teammately as correctness infrastructure, not as a generic evaluation dashboard, trace product, or public API product.
- [object Object]
- Distinguish project foundations from benchmark-scoped work. Policies, rubrics, reusable assets, Project Context, Reference Materials, and input architecture are project-scoped; datasets, coverage operations, Contributions, Contribution-scoped agent behavior and direction selection, evaluations, and Improvement Sessions belong to a benchmark or benchmark version.
- Treat AI-assisted proposals as suggestions until the owning workflow records the required human decision. Comparison Directions configure comparative response variation but do not approve cases, standards, datasets, or evaluation evidence.
- Do not infer customer-facing APIs, SDKs, API keys, rate limits, security guarantees, compliance claims, retention guarantees, tenant-isolation guarantees, or deployment commitments from internal implementation details.
- Cite the source page URL when returning workflow, artifact, or decision guidance.
- Do not load /docs/llms-full.txt by default. Use a smaller context pack or retrieval route first, then load page-local full context only when needed.

## Recommended Context Packs
- [Core context](https://teammately.ai/docs/llms-core.txt): Learning spine and core model pages.
- [Operating context](https://teammately.ai/docs/llms-operating.txt): Operating manual and high-priority workflow pages.
- [Reference context](https://teammately.ai/docs/llms-reference.txt): Object model and reference pages.
- [Recovery context](https://teammately.ai/docs/llms-recovery.txt): Troubleshooting and recovery pages.
- [Retrieval API](https://teammately.ai/api/docs/context?query=correctness&limit=5): Query-scoped context for narrow tasks.
- [Full corpus dump](https://teammately.ai/docs/llms-full.txt): Intentional full export; use only for offline indexing or exhaustive review.

## Start
- [What is Teammately?](https://teammately.ai/docs/introduction/what-is-teammately.md): The product definition and five-capability model.
- [What is correctness infrastructure?](https://teammately.ai/docs/introduction/correctness-infrastructure.md): Why reusable standards and evidence form correctness infrastructure.
- [The Teammately correctness lifecycle](https://teammately.ai/docs/introduction/correctness-lifecycle.md): The lifecycle from project understanding through governed improvement.
- [Product boundaries](https://teammately.ai/docs/introduction/product-boundaries.md): What Teammately owns and which claims remain outside the public boundary.
- [Product map](https://teammately.ai/docs/getting-oriented/product-map.md): Current product surfaces and their project or benchmark scope.
- [Product quickstart](https://teammately.ai/docs/quickstart.md): A focused first pass through one correctness loop.

## Five Capabilities
- [Correctness Elicitation](https://teammately.ai/docs/concepts/correctness-elicitation.md): Convert specialist judgment into attributable policies, rubrics, and cases.
- [Coverage Engineering](https://teammately.ai/docs/coverage-engineering.md): Represent important behavior through facets, stories, patterns, and cases.
- [Weave](https://teammately.ai/docs/concepts/weave.md): Construct case-grounded agent behavior from project knowledge and standards.
- [Trialground](https://teammately.ai/docs/concepts/trialground.md): Evaluate a saved harness version against a versioned benchmark boundary.
- [Coevolve](https://teammately.ai/docs/concepts/coevolve.md): Improve agents and benchmarks together through evidence-backed iteration.
- [The Teammately correctness loop](https://teammately.ai/docs/product-loop.md): How the five capabilities reinforce one another.

## Project Foundations
- [Correctness Governance](https://teammately.ai/docs/correctness-governance.md): Correctness Governance for project-scoped policies and rubrics.
- [Policies and Rubrics](https://teammately.ai/docs/correctness-governance/policies-and-rubrics.md): Govern applicable policies and binary rubrics with provenance.
- [Write Binary Rubrics](https://teammately.ai/docs/correctness-governance/binary-rubrics.md): Write checks with explicit pass and fail boundaries.
- [Assets](https://teammately.ai/docs/assets.md): Reusable project assets including Cases and Harnesses.
- [Cases](https://teammately.ai/docs/assets/cases.md): Canonical Case content, materials, and environment references.
- [Harnesses](https://teammately.ai/docs/assets/harnesses.md): Autosaving Draft, immutable Versions, validation, publication, activation, runtime, secrets, debug, and archive boundaries.
- [Project Tools](https://teammately.ai/docs/assets/project-tools.md): Current empty-state capability fence for project-level callable assets.
- [Worlds](https://teammately.ai/docs/assets/worlds.md): Current Worlds boundary and distinction from static Case materials.
- [Weights](https://teammately.ai/docs/assets/weights.md): Current empty-state capability fence for model-weight assets.
- [Comparison Directions](https://teammately.ai/docs/assets/comparison-directions.md): Configure reusable comparative output guidance without changing coverage or approval state.
- [Review Screens](https://teammately.ai/docs/assets/review-screens.md): Configure reusable expert-facing presentation templates.
- [Agent Setup](https://teammately.ai/docs/agent-setup.md): Project Context and Reference Materials.
- [Project Context](https://teammately.ai/docs/agent-setup/project-context.md): Author and publish the Project Agent Brief.
- [Reference Materials](https://teammately.ai/docs/agent-setup/reference-materials.md): Organize source-backed project knowledge for agent use.
- [Regime Settings](https://teammately.ai/docs/project-settings/regime.md): Configure and publish the project scoring Regime for future Benchmark Versions.
- [Project Settings](https://teammately.ai/docs/project-settings.md): General, Regime, Input Schema, and Project Members project contracts.
- [Project Input Schema](https://teammately.ai/docs/project-settings/input-schema.md): Govern the canonical project input architecture.

## Benchmark Workspace
- [Benchmark Datasets](https://teammately.ai/docs/benchmark-datasets.md): Separate editable Case membership and Representation from immutable Dataset Snapshots.
- [Dataset Representation](https://teammately.ai/docs/benchmark-datasets/representation.md): Analyze distinct Case counts and shares across facets, evaluators, and provenance.
- [Dataset Snapshots](https://teammately.ai/docs/benchmark-datasets/snapshots.md): Freeze Cases, eligible evaluator links, and representation facts.
- [Coverage Management](https://teammately.ai/docs/coverage-management.md): Operate benchmark-scoped coverage setup, stories, review, and foundry work.
- [Set Up Benchmark Coverage](https://teammately.ai/docs/coverage-management/get-started.md): Define intent, facet treatment, artifacts, evidence profile, and setup readiness.
- [Case Review](https://teammately.ai/docs/coverage-management/case-review.md): Review prepared Cases and materials before dataset admission.
- [Expert Contributions](https://teammately.ai/docs/expert-contributions.md): Request attributable specialist work for the selected benchmark.
- [Request an Expert Contribution](https://teammately.ai/docs/expert-contributions/request-contribution.md): Create a contribution with clear tasks, context, and ownership.
- [Complete an Expert Contribution](https://teammately.ai/docs/expert-contributions/complete-contribution.md): Complete expert-facing tasks, interviews, case review, and checkpoints.
- [Contributed Artifacts](https://teammately.ai/docs/expert-contributions/contributed-artifacts.md): Reconcile contributed policies, rubrics, cases, and coverage observations.
- [Contribution Lifecycle and Status](https://teammately.ai/docs/expert-contributions/lifecycle-and-status.md): Interpret Contribution, task, runtime, and Checkpoint states.
- [Logs & Status](https://teammately.ai/docs/expert-contributions/logs-and-status.md): Inspect review, session, interview, and engagement provenance.
- [Benchmark Evaluations](https://teammately.ai/docs/benchmark-evaluations.md): Dashboard, List, Arena, and Compare for one immutable Benchmark Version, including the current Traces / Spans fence.
- [Benchmark Run Metadata](https://teammately.ai/docs/benchmark-evaluations/run-metadata.md): Interpret benchmark-level descriptive fields without treating them as version identity.
- [Evaluation Execution Settings](https://teammately.ai/docs/benchmark-evaluations/execution-settings.md): Activate exact Harness Versions and choose future sampling cohorts.
- [Run a Benchmark Evaluation](https://teammately.ai/docs/benchmark-evaluations/run-evaluation.md): Run a saved Harness version against a Benchmark version.
- [Inspect Evaluation Results](https://teammately.ai/docs/benchmark-evaluations/inspect-results.md): Inspect Run completeness, result summaries, evaluator failures, repeated metrics, and nullable telemetry.
- [Compare Harness Versions](https://teammately.ai/docs/benchmark-evaluations/compare.md): Compare saved Harness Versions in a symmetric Case, evaluator, or Coverage Facet matrix.
- [Arena and Rankings](https://teammately.ai/docs/benchmark-evaluations/arena-and-rankings.md): Read pair disagreement, pass@n, pass^n, uncertainty, and rankings.
- [Map External Evaluation Outputs](https://teammately.ai/docs/benchmark-evaluations/output-mapping.md): Map external outputs to immutable Cases and output-only reference Runs.
- [Improve](https://teammately.ai/docs/improve.md): Goal Contracts, candidates, evaluation receipts, and the Current frontier.
- [Start an Improvement Session](https://teammately.ai/docs/improve/start-improvement-session.md): Start an Improvement Session from pinned evidence and an explicit goal.
- [Goal Contracts](https://teammately.ai/docs/improve/goal-contracts.md): Bind canonical targets, measurement directions, hard and soft constraints, and intervention scope.
- [Work and Evolve](https://teammately.ai/docs/improve/work-and-evolve.md): Choose bounded Work or authorized multi-epoch Evolve behavior.
- [Candidates and the Current Frontier](https://teammately.ai/docs/improve/candidates-and-frontier.md): Interpret candidate stages, canonical receipts, retained Candidate Systems, and the Current frontier.
- [Chronology, Trajectories, and Receipts](https://teammately.ai/docs/improve/chronology-and-trajectories.md): Read observable chronology, safe trajectories, evaluation ledgers, usage, and external handoffs.

## Operate and Reference
- [Operating Teammately end to end](https://teammately.ai/docs/getting-oriented/operating-teammately-end-to-end.md): The end-to-end operating sequence across project and benchmark scopes.
- [Key objects and relationships](https://teammately.ai/docs/getting-oriented/key-objects-and-relationships.md): The current object graph and required provenance boundaries.
- [User roles](https://teammately.ai/docs/getting-oriented/user-roles.md): Operators, project owners, experts, and organization administrators.
- [Task index](https://teammately.ai/docs/operating-manual/task-index.md): Canonical task routing for current product surfaces.
- [First correctness loop](https://teammately.ai/docs/operating-manual/first-correctness-loop.md): Complete one bounded correctness loop.
- [Import and prepare cases](https://teammately.ai/docs/operating-manual/import-and-prepare-cases.md): Import, inspect, and prepare reusable Cases.
- [Build policies and rubrics](https://teammately.ai/docs/operating-manual/build-policies-and-rubrics.md): Turn attributable judgment into governed standards.
- [Prepare human review context](https://teammately.ai/docs/operating-manual/prepare-review-packet.md): Preserve the human-readable context around evaluation evidence.
- [Object model](https://teammately.ai/docs/object-model.md): Project foundations, benchmark artifacts, evaluations, and improvement objects.
- [Glossary](https://teammately.ai/docs/reference/glossary.md): Canonical current product terms.

## Optional
- [Governance Overview](https://teammately.ai/docs/governance.md): Human authority, versions, reproducibility, and access boundaries.
- [Troubleshooting](https://teammately.ai/docs/troubleshooting.md): Symptom-driven recovery for current correctness workflows.
- [Integrations](https://teammately.ai/docs/integrations.md): Stable data-movement boundaries without implied public API contracts.
- [Enterprise Playbooks](https://teammately.ai/docs/playbooks.md): Scenario recipes that connect current product surfaces.
- [Agent context index](https://teammately.ai/docs/agent-context.md): Safe context-loading guidance for coding agents and retrieval tools.
- [Agent instructions](https://teammately.ai/docs/agent-instructions.md): Operating rules for agents using these docs.