Project Harness

Purpose

Use this domain for the project's static operating surface: entrypoints, core-change evidence, diagnostics, design contracts, fast feedback, acceptance, change safeguards, and recovery boundaries.

Load When

• A project evidence pass needs static repository, history, diff, or core-path context.
• Agent Work Loop Project Harness needs reproducible startup, supported operation, feedback, observability, acceptance, approval, or recovery evidence.
• A report, hook, or review flow needs sensitive-code handling.
• Git, CI, or local lifecycle checks need review-trigger guidance.
• Harness-analysis output needs to separate static presence, configured policy, and observed execution evidence.

Owns

• agent-verify-loop.md: cross-stack verification-harness framework for self-verifying agent loops (service chains, UI, mini-programs, pipelines, model output).
• core-change-watch.md: project profile, history, core-path, diff, and recommended-read evidence.
• observability.md: diagnostic routes and correlated runtime evidence.
• project-overlays.md: project-type additions to the common evidence model.
• design-md-contract.md: design source-of-truth boundaries.
• ui-and-system-drivers.md: choosing where the agent injects control into the system under verification (D0..D4 drive planes), and the separate selection of drive and observation adapters.
• friendly-cli.md: agent-friendly command contracts.
• git-hooks.md: Git hook lifecycle placement and evidence.
• recovery-evidence.md: rollback and recovery presence, wiring, safe exercise, applicability, and validation evidence.
• review-trigger.md: recommendation and review-trigger policy.
• sensitive-code.md: sensitive-code review and reporting boundaries.
• sensitive-write-boundary.md: pre-write confirmation gate for project-control surfaces.
• verification-environment.md: claim-driven construction and calibration of mock, emulated, ephemeral, and sandbox verification environments.

Does Not Own

• Agent instruction, host asset, or provider feature taxonomy; use ../agent-customize/routing.md.
• Session-derived Task Episode evidence; use ../session-evidence/.
• Combined Agent Work Loop evidence mapping; use ../../models/agent-work-loop.md#construct-task-episode-evidence.
• Repeated-work owner selection; use ../loop-engineering/.

Read Next

• Use agent-verify-loop.md for designing a self-verifying regression harness around an agent loop.
• Use ui-and-system-drivers.md when more than one driver class could exercise a UI, desktop, mobile, or terminal case and the injection point must be chosen by claim.
• Use review-trigger.md for recommendation and review-trigger policy.
• Use core-change-watch.md for static project evidence and changed-core routing.
• Use observability.md for diagnostic coverage and runtime evidence.
• Use project-overlays.md for scoped project-type evidence additions.
• Use design-md-contract.md for DESIGN.md ownership and verification.
• Use friendly-cli.md for command interface design.
• Use recovery-evidence.md for rollback and recovery inspection.
• Use sensitive-code.md for sensitive-code review boundaries.
• Use sensitive-write-boundary.md for pre-write confirmation on project-control surfaces.
• Use verification-environment.md when the real verification environment is unavailable or impractical and the agent must choose a credible substitute.
• Use git-hooks.md for Git hook lifecycle placement and evidence.