Purpose

Valdris SDLC Harness turns agent engineering work into a reviewable run packet. It keeps coding agents external, but gives them a repo-specific operating manual, bound workload classification, Layer 0 foundation assurance, workflow nodes, proof gates, and handoff rules.

The lifecycle control flow is commission -> route-goal -> assure -> connect-runtime -> execute-workflow -> prove-govern -> trust-improve. These seven skills select the owning Valdris system. Inside routed work, a separate eight-skill catalog selects one primary workflow skill per intake, delivery, or proof-handoff phase. The catalogs are complementary, not a fifteen-way flat choice.

Lifecycle skills, workflow skills, and commissioned lanes have different jobs. A lifecycle skill owns one control-plane system end to end. A workflow skill selects the phase procedure. A lane loads project-specific context, ownership, commands, runtime/model policy, and gate emphasis. The immutable route may bind the work skill and lane while the lifecycle registry governs the surrounding system flow; none is a substitute for another.

A commissioned target must review and commit its generated .valdris-harness pack and front-door files before it creates an immutable route. Routing then runs from the target root with the committed target-local router. Independent review covers the frozen evidence set after all applicable technical evidence, including Red Zone authorization and smoke proof; changing that set requires a fresh review.

Agent Entry Points

Use Engineering Task Routing for normal tasks. Use Layer Zero Assurance to bind classification and resolve foundation requirements before feature or architecture implementation. Use GitNexus Code Intelligence before code-impacting claims. Use Production Readiness 13 Domains when production behavior could be affected.

The thirteen production domains are the current shared assurance baseline, not an exhaustive list of literal system layers. Asynchronous orchestration is cross-cutting and must be evaluated across every domain it affects.

Authoritative Repo Files

• README.md for the product overview.
• AGENTS.md and CLAUDE.md for runtime front doors.
• controls/workload-taxonomy.v1.json and scripts/workload-classification-gate.mjs for bound workload classification.
• controls/foundation-layer.v1.json and scripts/foundation-gate.mjs for Layer 0 assurance.
• skills/registry.json and skills/codex-routing.yaml for the seven lifecycle and eight workflow skill catalogs.
• scripts/route-lifecycle-skill.mjs for deterministic lifecycle-system routing.
• scripts/commission-harness.mjs for generated pack structure.
• scripts/verify-harness.mjs for adversarial proof coverage.
• docs/HARNESS_REPO_MAP.md for the deeper file-by-file map.