[ITSGJY]

Summary

• expose the existing typed project_build lifecycle directly on the Adaptive Runtime surface at rank 94, immediately before project_validate
• preserve the canonical project_build_v1 Runner capability, Standard approval, authority, typed planning, durable Job identity, Runner-owned execution, and observe_jobs continuation
• keep same-durable-Job continuation guidance in the default compact descriptor so a pending build is observed rather than redispatched
• add tool-contract and MCP surface regressions plus a build-specific measurement record

Measurement

Using the evidence discipline established by yyjeqhc#959, two fresh A/B pairs were run in opposite order with the same Codex Host, model, prompt, Runner, and Rust fixture. The comparison changed only model-facing exposure of the existing canonical build lifecycle.

• compact surface: 52,762 -> 54,819 serialized tool bytes (+2,057 / +3.899%)
• gateway path: read_tool_manifest x2 -> call_runtime_tool(project_build)
• Direct path: one project_build call
• completed Host MCP calls: 3 -> 1
• model responses: 8 -> 5
• input tokens: 145,619 / 141,884 -> 81,289 / 81,255
• median wall time: 20,895.5 ms -> 10,772 ms
• correctness: 2/2 successful builds on both variants, build artifact present, source workspace clean, zero ActionAudit failures

Input-token and wall-time values are supporting observations for this fixed Host/task, not a general latency claim. The full method and evidence boundaries are recorded in docs/experiments/project-build-direct-measurement.md.

Contract

This is a model-facing discovery change. Build planning, Runner execution, authority, approval, Job identity, and result projection keep their existing canonical paths. Long builds remain the same durable Job, with observe_jobs used for continuation; a pending result does not authorize retry or redispatch.

Validation

• webcodex-tool-contracts: 289 passed
• compact project_build continuation-guidance regression: passed
• Adaptive MCP Direct exposure regression: passed
• cargo fmt --check
• git diff --check
• Codex committed-diff review: no actionable regressions
• agy committed-diff review: no findings
• fork CI 37575588180: success on 44225edbba1b7fbb8e5f82f5048b6831678a7702
• fork code scanning 37575598119: success on the same head

Related to yyjeqhc#599.

Upstream submission: yyjeqhc#960.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[ITSGJY]

Upstream submission yyjeqhc#960 has been merged.