Skip to content

task: R01 — retire legacy root paths and close core dependencies #412

Open

Description

@Dong1017

Current delivery decision (2026-11-15)

Split legacy-path retirement and actual dependency closure into three bounded implementation PRs. Retire already-migrated old paths early instead of keeping every shell until the last integration change. Moving implementation, migrating active consumers/docs/tests and deleting old root modules belong in the same PR. Overall isolation is a real-tree result, not an issue-status or merge inference. This issue closes root-level scattering, old entrypoints and core dependency boundaries; it does not promise internal refactors of existing subpackages.

Related: #387, #371, Apps migration #408, shared contracts/resources #451, and Apps-only MindIE relocation #452.

PR checklist

  • R1 — Retire existing legacy root paths and finish canonical inspection binding. Delete the seven Knowledge shells (_skill_selection, _source_snapshot, catalog, knowledge_operations, knowledge_store, research, skill_catalog), the three B01 root paths (evaluation, execution, reporting), and root inspection. In this same PR update all active consumers, including the CLI/arguments if they have not moved yet, tools/current tests/docs, the manifest and affected protected-path rules. Root inspection is not a trivial alias: first preserve its command/snapshot/repository-identity binding in the canonical Apps composition at apps/inspection.py over the existing infra facts API. Do not delete behavior or injection assertions. Maintain the shrinking exact root-file inventory introduced in C1. If R1 is scheduled earlier, record its removals so that C1 initializes the guard from the real tree; R1 does not need a second root guard. When catalog is retired, consumers take REPO_TYPES from resources.repository_packs as the same object (identity assertion), never a copy. Record the exact remaining compatibility pairs at each step: the eight root-shell pairs disappear with the seven shells, and the neutral-leaf importer (infra.inspection -> resources.repository_packs after C2) must be covered by a permanent rule outside compatibility_pairs, with a negative test that only that exact edge is allowed. Leaving it in compatibility_pairs keeps allowlist: 1 in first_wave_report() and makes the final blockers == () unreachable.
  • R2 — Close artifact-primitive dependencies in six Harness modules. Remove the infra.artifacts reachability from agent_sessions, candidates, optimization, optimization_session, project and task_protocol, using equivalent Harness-owned primitives or explicit suitable ports. Preserve writer/digest/hash/lock bytes, atomicity, failure and platform semantics. The existing report has 12 path hits from these six modules, not 12 modules: each module carries both the package-level inferrsi.infra edge and the inferrsi.infra.artifacts edge, so removing both is the completion bar; replacing them with imports of other inferrsi.infra modules does not count. The same PR also verifies that Harness has no inferrsi.model edge (the current tree has none); the final gate counts both inferrsi.infra and inferrsi.model.
  • R3 — Close five real execution boundaries and prove the combined tree. Handle harness/task_protocol.py, the moe-forward-npu benchmark/grader, qwen21-denoise-npu benchmark and qwen3-tts-npu benchmark through legitimate bounded execution/host interfaces. Preserve candidate/sandbox/integrity controls and frozen input semantics. This PR contains real boundary fixes and final integration acceptance; do not create a ninth verification-only PR.

Dependencies and merge discipline

R1 has no hard prerequisite on C1/C2 (#451) or A1 (#408): deleting existing legacy paths does not itself require relocating a neutral resource leaf first. Prefer C1 → C2 → R1 in the serial queue to avoid rewriting contract/resource policy twice; do not wait for A1/A2/M1 to retire the existing eleven paths. C2 does require the contracts package delivered by C1. Start the shrinking root-file inventory guard in C1, then update its exact set here and in every removal PR. The final two-file target is checked only after the actual remaining moves.

R2 is logically independent and may be scheduled earlier if useful, but only one Claude implementation session runs at once. R2 and R3 both touch task_protocol; R3 must integrate R2's actual head. R3 implementation planning can start earlier, but its final integration/tested head must contain all accepted C/A/M/R predecessor changes. Do not represent a planned DAG as authorization for parallel coding sessions.

Acceptance

  • R1 really removes the eleven paths, updates all active consumers in that PR and cleans the corresponding compatibility entries, recording the allowlist count at each step and which PR removes the last pair. Canonical injection behavior remains covered; frozen historical evidence bytes are not edited.
  • R2 eliminates the six modules' infra/model reachability with positive/negative and platform-sensitive checks, covering both the package-level inferrsi.infra and inferrsi.infra.artifacts edges, not a permanent allowance for Harness-to-infra imports, and keeping Harness -> model at zero.
  • R3 resolves the actual five reviewed execution-loader boundaries. Do not merely remove declarations, relax scanners or change expected output to manufacture isolation.
  • On the final exact tree: first_wave_report().blockers == (), new_core_closure_isolated is True, and isolation_complete is True. Harness has zero edges into inferrsi.infra or inferrsi.model; the permanent neutral-leaf importer rule from R1 is part of this gate and is not counted as a compatibility pair. Negative controls reject old core paths, hidden dynamic imports and temporary forwarding.
  • The tracked root-file set is exactly {__init__.py, __main__.py}; wheel checks separately allow only the declared generated _version.py metadata. Installed-wheel/resources/host and Windows/Linux checks pass without changing original byte/fact/security contracts.
  • Record all retired paths and replacements in CHANGELOG. Root physical cleanliness and dependency isolation are both required; neither substitutes for the other.

Delivery discipline

  • One implementation Claude session handles only the current PR in an isolated worktree. A second Claude session may only plan the next PR in a separate planning worktree; it must not edit source.
  • Move implementation, update all active consumers/docs/tests, and delete the corresponding old root modules in the same PR. Do not introduce permanent compatibility shells or module-registry aliases.
  • Preserve CLI command behavior, original evidence bytes, record/format contracts and isolation/security checks. Announce retired Python import/serialization module paths; historical frozen evidence remains pinned to its original commit or wheel, not rewritten.
  • Before each implementation or review handoff, refresh main and confirmed dependencies. A completed plan is not implementation or approval. No automatic merge.

Historical context retained

The 2026-11-12 R01 scope already required retirement of seven Knowledge shells, three B01 root forwarders, the legacy manifest/allowlist and affected protected roots; installed-wheel/host checks and real positive/negative closure remained mandatory. Its final dependency list included K01/H02/B01/S01/OBS/A-entry, with no mandatory tag wait. The earlier permanent root CLI-forwarder carve-out and the assumption that every retirement must wait for the final PR are superseded by the owner's revised delivery decision. The neutral resource leaf still requires an explicitly reviewed no-core-import rule at its new canonical location; infra must not become a fourth core. No new product feature, K02 resource re-migration or formal replay/model experiment is included. Earlier issue discussion is retained; this update does not claim any new review or implementation acceptance.

Activity

  1. added
    kind:taskBounded delivery task with a delegated slice
    area:processRepo-wide process, governance, privacy and security rules
    on Nov 12, 2026
  2. amy-why-3459 commented on Nov 12, 2026

    @amy-why-3459
    Collaborator

    谢谢,R01 作为收尾任务,验收设计得很好:"真实树 isolation_complete is True"加上"加回旧叶子或转发即失败"的反例。补充几点:

    1. "兼容期结束证据"请给出可检查的定义。 例如在合入前用 rg/import 扫描证明 src、tools、tests、docs 以及三个宿主插件里都没有旧根路径的引用;外部消费者的情况也请在 PR 里说明。正文说"不额外等待固定 tag 周期",那么 CHANGELOG 里"移除路径 → 替代路径"的对照表就是用户唯一的迁移指引,建议逐条列出。
    2. 两条永久规则(infra.inspection → _repo_pack_data 技术叶子消费者,以及根 cli/arguments/__main__ 精确转发)建议在门禁里各写一个反例测试,证明规则只放行这一条边或这一种形态,不会变成通配。
    3. 前置顺序。 R01 依赖 K01、H02、B01、S01、OBS、A-entry,其中 A-entry、OBS、S01、B01 都还依赖 Publish standardized Apps run output #397 合入和 task: Apps 应用层详细设计与迁移计划(CLI / Python / Plugin) #371 接受(task: Apps 应用层详细设计与迁移计划(CLI / Python / Plugin) #371 目前仍是 status:proposed)。建议在 rfc: K/H/B 与 Apps 公开接口详细设计(最小交接与模块边界) #387 或本 issue 里维护一张依赖状态表,让 R01 的启动条件一目了然。
  3. changed the title [-]task: R01 — 退役兼容壳与 allowlist,完成整仓隔离[/-] [+]task: R01 — retire legacy root paths and close core dependencies[/+] on Nov 15, 2026
  4. amy-why-3459 commented on Nov 15, 2026

    @amy-why-3459
    Collaborator

    看到正文改成了 R1/R2/R3 三个 PR 的拆分。我在当前 main(1373ef8e)上跑了 first_wave_report(),和新写的内容逐项对照如下:

    1. blockers 共 4 项,能和新拆分对上。 分别是 allowlist: 9 compatibility pairs、legacy run forwarders: 3(对应 R1 的 evaluation / execution / reporting)、harness reaches model/infra ... (12 hits)(对应 R2)、runtime resource loaders: 5(对应 R3)。R1 删除 11 个路径,但 allowlist 只有 9 对。建议在 R1 验收里写明 R1 之后 allowlist 应清空,或者列出剩下的确切几对及其负责 PR。
    2. R2 的 12 处命中。 实际是 6 个模块各有两条边:inferrsi.infra 和 inferrsi.infra.artifacts。这和正文"12 path hits from these six modules"一致。所以 R2 必须连包级的 inferrsi.infra 边一起消除,改成 import infra 下的其他模块不算完成。另外,汇总字符串只显示第一个模块(agent_sessions -> inferrsi.infra (12 hits)),review 时建议看 harness_model_infra 的完整列表。
    3. R1 删除 catalog 会影响 REPO_TYPES。 arguments.py 现在从 .catalog 导入 BACKENDS / KNOWLEDGE_REPOS / KNOWLEDGE_SCOPES / REPO_TYPES,而 REPO_TYPES 实际定义在 _repo_pack_data.py。如果按首选顺序在 C2 之后做 R1,消费者应直接从 resources.repository_packs 取 REPO_TYPES,保持同一个对象(和我在 task: consolidate shared contracts and packaged-resource access #451 提的 identity 检查一致)。如果 R1 提前,也不要在 Knowledge 的 canonical 模块里复制一份元组。
    4. 我上次提的两条永久规则:根 CLI 转发这条已被新决定取消,不再需要。infra.inspection → _repo_pack_data 这条在 C2 后要迁到新位置。建议在 C2 或 R1 里配一个反例测试,证明它只放行这一条边。
  5. Dong1017 commented on Nov 15, 2026

    @Dong1017
    CollaboratorAuthor

    Posted by an agent on behalf of @Dong1017 (incremental review only; points already raised by Amy are not repeated).

    Ran first_wave_report() on current main d32eb968; the four blockers match Amy's numbers exactly. Two additions to her 6892341854 point 1:

    1. Exact mapping of the 9 compatibility_pairs. Eight are removed by R1 because they are edges from the seven root shells being deleted: _skill_selection -> knowledge._skill_selection, _source_snapshot -> knowledge._source_snapshot, catalog -> knowledge.catalog, catalog -> _repo_pack_data, knowledge_operations -> knowledge.knowledge_operations, knowledge_store -> knowledge.knowledge_store, research -> knowledge.research, skill_catalog -> knowledge.skill_catalog. The ninth, infra.inspection -> _repo_pack_data, is not a root-file edge: src/inferrsi/infra/inspection.py:10 imports the neutral leaf and :56 calls it. Even if C2 rewrites that pair to the new resources.repository_packs location first, R1 would leave allowlist: 1 compatibility pair, not zero. (The three legacy run forwarders are a separate blocker and are cleared by R1; they are not among the nine pairs.)

    2. That last pair cannot stay in compatibility_pairs if R3's acceptance is blockers == (): tests/_gate_policy_support.py:553-554 counts every manifest pair as allowlist: N, and removing the pair while the import remains produces violations: 1 — inferrsi.infra.inspection is a second importer of inferrsi._repo_pack_data (tests/_gate_policy_support.py:489-494; reproduced by calling first_wave_report() with only that pair removed). So reaching the final zero also needs a permanent neutral-leaf importer rule outside the compatibility allowlist (with the negative test Amy asked for), or an equivalent gate change; the acceptance text should name that mechanism, otherwise R1 leaves one blocker that no code deletion in the checklist can remove.

  6. zengchuang-hw commented on Nov 17, 2026

    @zengchuang-hw
    Contributor

    独立源码检视(普通 Issue,不是 PR 批准):基于 main 61303fa8150472dfbda62ec8d77908bc05c73f3b。结论:根路径退役已有实际实现,但不能据此认定“core dependencies 已闭合”或关闭本 Issue。

    已核实的实现

    • 当前 src/inferrsi/ 源码根文件仅 __init__.py、__main__.py;后者与 console script 均走 inferrsi.apps.cli。R1 的 11 个旧根模块已删除,canonical 实现在 Knowledge / Apps;不是仅改目录说明。
    • tests/test_package_root_inventory.py 对额外根文件和已退役名称回归有反例;tests/test_knowledge_compatibility_shells.py 检查旧 import / pickle 路径失败及不留 sys.modules alias。CHANGELOG.md:114–146 明确旧 Python import / pickle 兼容性中断和历史 evidence 不重写。这是实现与迁移说明的证据,不等于完整接口验收。

    仍然存在的闭合缺口

    1. Harness 仍直接依赖 Infra。例如 agent_sessions.py:13、task_protocol.py:11 引用 infra.artifacts。在该 SHA 实跑 first_wave_report() 得到 12 条可达路径,涉及 agent_sessions、candidates、optimization、optimization_session、project、task_protocol。这里的 12 是图路径命中数,不是 12 个不同的直接 import。
    2. 五个 runtime resource loader 仍在源码与 gate 中。例 task_protocol.py:14–21 用动态 module spec 执行 task 源码;其余位于 moe-forward-npu benchmark/grader、qwen21-denoise-npu benchmark、qwen3-tts-npu benchmark。当前 gate:836–843 正确把这些声明保留为 blocker,而非视为已隔离。实测 violations=(),但 new_core_closure_isolated=False、isolation_complete=False。因此“scan 无 violations”不能替代闭合验收。

    后续闭合证据应绑定实际交付 SHA,证明上述依赖/执行边界按确认过的契约被处理,同时保留 artifact bytes、原子写/锁、候选 sandbox 与完整性负例;不能通过删除扫描项、放宽声明或增加永久 alias 让 gate 变绿。这不是把动态 task 执行本身定性为漏洞,也不是要求直接删除它。

    本次验证范围
    在独立源码目录用 Linux / Python 运行 root inventory、target-module import gate、legacy-run forwarder gate、knowledge compatibility shells、evidence paths、repository-packs facade 六个测试文件:128 collected,127 通过,1 跳过(archive 无 Git metadata 的 tracked inventory 项);另实跑上述 isolation report。未运行全套测试、installed-wheel/resource 验证、真实 native-host 或 Windows 验证,未验证完整 CLI bytes / frozen-evidence 跨版本等价,不能替这些项目盖章。

    本次没有读取或引用其他人的评论/Review 结论,也没有将 Issue 状态或合并状态当作通过依据。需求版本方面,仅按可隔离的“退役旧入口并真实闭合依赖、保留接口/证据/安全边界”范围检视;没有确认混有决策转述的修订段落所述分拆、顺序例外或最终验收谓词,故本意见不表示接受整套架构、不承诺实施,也不是对本 Issue 全部验收条件的 PASS。

  7. Dong1017 commented on Nov 17, 2026

    @Dong1017
    CollaboratorAuthor

    Posted by an agent on behalf of @Dong1017. This is a proposed R3 design for discussion before implementation; nothing below is implemented, and it does not claim review acceptance.

    R3 design: close the five execution-loader boundaries

    Read against the R2 head 1d0190e01ea59954a679054e697a22450edce011 (#478, not merged yet). The implementation will branch from main after R2 merges, so its tested head contains R2.

    Problem

    first_wave_report() keeps one blocker after R2: runtime resource loaders: 5. Each is an importlib.util.module_from_spec call that the gate admits through the temporary RESOURCE_LOADERS table in tests/_gate_policy_support.py:

    # Site What it loads Why it loads by path
    1 harness/task_protocol.py protocol() The task's frozen optimization.py (validate_arguments, arguments, assess), into the Harness process Task bundles are not importable packages
    2 moe-forward-npu/benchmark.py, module scope Sibling processes.py (MARKER, cleanup) Sibling of a non-package directory
    3 moe-forward-npu/grader.py load() task/baseline.py and candidate/implementation.py from the run directory, inside each torchrun rank The rank must call compute in-process to time it
    4 qwen21-denoise-npu/benchmark.py load() workload, reference, baseline and candidate from output/snapshot/ The evaluator must call compute in-process to time it
    5 qwen3-tts-npu/benchmark.py, module scope Sibling evaluation.py Sibling of a non-package directory; also runs under python -I

    #412 requires legitimate bounded execution/host interfaces that keep candidate, sandbox and integrity controls and frozen-input semantics, and a final tree with blockers == (), new_core_closure_isolated is True and isolation_complete is True. Removing declarations, relaxing scanners or editing expected output alone does not count.

    Decision

    The sites fall into three kinds. Each kind gets one treatment.

    Kind A, site 1: Harness runs the task protocol as a child process

    Harness stops executing task code in its own process. task_protocol runs the frozen protocol file as a program, the same way Harness already runs benchmark.py:

    • Command: sys.executable -I -B <task root>/optimization.py. The interpreter is the one that runs the protocol today. -I keeps PYTHONPATH, user site and the current directory out; -B together with script execution means the frozen source is compiled every time and no bytecode cache is written or read, which is the existing "always read the frozen source" rule.
    • Request: one JSON object on stdin, {"operation": ..., "arguments": {...}}, for validate_arguments, arguments and assess. Paths travel as strings and the protocol turns them back into Path.
    • Response: one JSON object on stdout, either {"result": ...} or {"error": {"type": <exception class name>, "message": str(error)}}. stderr passes through to the caller.
    • Harness raises TaskProtocolError (a ValueError) that carries the child's exception type, so the recorded reason stays "<Type>: <message>" (pinned today by test_assessor_exception_preserves_type_and_settles_failed_attempt). An argparse usage error in the child keeps exit status 2 for session start. A non-zero exit or malformed response without an error object fails fast.
    • The protocol contract becomes: a single standard-library file that is also a program (if __name__ == "__main__": dispatch). qwen21-denoise-npu/optimization.py and the test fixture protocols gain that dispatch; their functions do not change.
    • harness/task_protocol.py changes, so the runtime identity refuses to resume sessions started before this change, as for any edit to it.

    Result: Harness has no dynamic import and no in-process path to task code. The CLI and Harness process never execute bundle code; bundle programs run only in child processes.

    Kind B, sites 2 and 5: remove the sibling load

    These load a trusted sibling of the same bundle only because the directory is not a package. The bundle already uses plain sibling imports elsewhere (oracle.py, the grader, the TTS baseline.py), but those rely on the script directory being on sys.path, which fails under python -I and for in-process test loading. Merging removes the need for any import path:

    • moe-forward-npu: move the 80 lines of processes.py unchanged into benchmark.py and drop it from TASK_FILES. Tests that used moe._processes use moe.
    • qwen3-tts-npu: rename evaluation.py to benchmark.py and move the 98-line CLI into it; the prepared bundle loses evaluation.py. Tests that loaded evaluation load benchmark. docs/tts-benchmark.md and docs/engineering-benchmarks.md follow. policy.json pins grader_sha256 to the grader file and runtime_sha256 to every bundle file, so a bundle prepared before R3 is refused by load_prepared afterwards, as after any grader edit today; it must be prepared again. Retained results are not rewritten.
    • moe-forward-npu: each run directory pins its own TASK_FILES in identity.json and is adjudicated in the same run, so no earlier run directory is reopened.

    No loader, sys.path change or sys.modules change replaces them.

    Kind C, sites 3 and 4: keep the two evaluator loaders under a permanent rule

    An evaluator must call the candidate's compute in the same process to time it, so some code must load the candidate file there. A static import cannot do this correctly: it caches by module name in sys.modules (a second load returns the first candidate), registers the trusted workload/reference/baseline modules where the candidate can import and patch them, and needs a sys.path change. Moving the call to runpy or exec only hides it from the scanner. So these two loaders stay, with the contract made explicit and tested:

    • They exist only in bundle program files (benchmark/tasks/<bundle>/, a directory without __init__.py), which run only as child processes.
    • Each loads only files inside its own run directory that are recorded in identity.json before execution and re-verified afterwards (the grader by _adjudicate, qwen21 by main() and the protocol's assess).
    • Each call creates a fresh module that is never registered in sys.modules; the candidate module must define a callable compute.

    The gate moves them from the temporary RESOURCE_LOADERS table, which counts as a blocker, to a permanent EVALUATOR_LOADERS rule with exactly these two entries, the same (function, spec variable) shape check, and no blocker. This follows the R1 precedent for the neutral-leaf importer: an irreducible edge covered by a narrow permanent rule outside the blocker list, with negative tests.

    Gate after R3

    • RESOURCE_LOADERS and the runtime resource loaders blocker are removed. A module_from_spec in harness/task_protocol.py or anywhere outside the two evaluator entries is an undeclared dynamic import and a violation.
    • EVALUATOR_LOADERS admits one call in moe-forward-npu.grader.load and one in qwen21-denoise-npu.benchmark.load. The gate also fails if a declared module is not a bundle program file.
    • new_core_closure_isolated no longer depends on loaders.
    • Expected real-tree result: violations == (), harness_model_infra == (), blockers == (), new_core_closure_isolated is True, isolation_complete is True.

    Negative controls (new or rewritten): a returning loader in harness/task_protocol.py; a second call or a call in another function of an evaluator; the same shape in another bundle file (workload.py), in a non-bundle Benchmark module, in Apps and in Harness; a declared module given an __init__.py; import_module and __import__ next to a declared loader. The existing negative controls for old root paths, forwarders and Harness → Infra/Model edges stay.

    Semantics to preserve and how they are checked

    Contract Check
    Protocol results are identical (settings, argv, assessment) Existing protocol tests run against the child-process path; JSON round trip of every returned value
    Protocol failure reason "<Type>: <message>" and argparse exit 2 Existing reason test plus a child-raised TypeError and a usage error
    Frozen source, no bytecode cache, protocol read from <session>/frozen Test that a planted __pycache__ is neither read nor written; -I excludes a planted PYTHONPATH module
    Harness never imports or execs task code Gate plus a test that the CLI process has no task module after a round
    MoE cleanup, marker and process-group behavior unchanged Existing test_moe_benchmark.py against the merged file
    TTS prepare/serve/collect/assess unchanged, standalone under python -I Existing TTS tests; the -I --help test unchanged
    Evaluator loads exact frozen files into fresh, unregistered modules New tests for both loaders: wrong path rejected by identity, sys.modules unchanged, two loads give two modules

    Platforms: H200 Linux full suite on candidate and unchanged main with a failed-id diff; native Windows for the protocol subprocess (test_optimization_session.py, test_project_optimization.py, test_qwen_optimization_protocol.py); GitHub full-check, installed-package and host-compatibility. No NPU run: the NPU paths are unchanged except the loaders' surroundings, and the qwen21 CPU --check-only path exercises its loader.

    Not claimed

    • No candidate sandbox: an in-process candidate can still reach the evaluator's objects, as today. The rule admits execution of frozen inputs; it does not isolate them.
    • No NPU performance or correctness result.
    • runpy and exec remain outside the gate's scan scope, as documented; R3 does not move any loader to them.

    Point for discussion

    Kind C is the only place where the gate admits code permanently. It is chosen as the final form, not a temporary step, so it does not need a later rework PR. The alternative is to keep the two evaluator loaders as blockers. That makes blockers == () unreachable unless the evaluators are deleted, so #412's final acceptance would have to be reworded.

    R3 implements Kind C as described unless a reviewer shows a bounded interface that removes in-process candidate loading without the static-import problems above. Such a counterexample reopens this point before the R3 PR is opened. Comments on kinds A and B, the gate rule or the negative controls are equally welcome before implementation starts.

  8. zengchuang-hw commented on Nov 17, 2026

    @zengchuang-hw
    Contributor

    Independent design-only increment for the proposed R3 boundary in comment 6948276105. This is not implementation acceptance, authorization to implement, or agreement to relax Issue #412's acceptance gates.

    Baseline: current main a02ef1863e29de4510272e65ae69341b1ef83f22. Previous own inspection: #412 (comment) (main 61303fa8150472dfbda62ec8d77908bc05c73f3b). I used the isolated proposal scope and inspected executable code without other reviewers' opinions or author result narratives. The referenced PR478 head was not treated as integrated or as this baseline.

    Current implementation versus proposal. harness/task_protocol.py:protocol still creates a module and executes optimization.py inside the caller process; prepare invokes its validate_arguments, and optimization_session.py:36,54 invokes arguments and assess. The current gate still declares five RESOURCE_LOADERS. My direct gate run returned zero violations but five runtime-resource-loader blockers plus 12 Harness→Infra hits; both closure/isolation flags are false. Moving execution to a child is therefore a substantive boundary change, not an already demonstrated property.

    Design conditions to resolve before implementation acceptance:

    1. The JSON child protocol needs a versioned operation-specific request/response contract, bounded stdout/stderr and an explicit deadline/termination policy. Preserve distinctions among malformed JSON, duplicate/nonfinite values, invalid schema, child exit2, nonzero exit, and transport failure. Stdout must be exclusively the envelope; diagnostics belong on stderr. Cover all three operations at their actual callers, not merely a standalone echo fixture. Define cleanup for child descendants as well as the direct child.
    2. -I -B addresses interpreter import/cache behavior; it is not a filesystem/network/security sandbox and does not itself pin file bytes. Freeze and verify the exact protocol/bundle identity before execution, reject incompatible resume identities, and test mutation/replacement cases. Post-execution verification detects drift but cannot undo side effects or guarantee unchanged bytes were executed.
    3. A permanent evaluator-loader exception is a policy change, not proof of stronger isolation. Preserve exact path/function/call-count restrictions and reject all additional dynamic-import sites; add the proposed adversarial controls before removing the old blocker. A passing scanner must not become tautological because the exception and expected result changed together. Distinguish static dependency closure from runtime execution isolation in the resulting report.
    4. Current MoE grader.load checks callable compute; current Qwen21 benchmark.load returns its module without that check. Neither loader signature itself accepts an expected digest. Callable/identity guarantees therefore need verification across each actual loading caller; do not infer them from the exception's name or module freshness alone. Keep imported-module registration, wrong-path, extra __init__.py, adjacent imports and repeated-load freshness as separate negative controls.

    Checks actually run: tests/test_target_module_import_gate.py and tests/test_bundle_manifest.py: 32 passed, 43 subtests passed, one pytest-plugin deprecation warning. A separate current-tree gate invocation produced the values above. AGENTS/review guidance is unchanged and was checked. These CPU/static checks validate the existing policy behavior, not the unimplemented R3 protocol or the proposed final gate state.

    Not run: candidate-versus-main full suites/failure-ID comparison, native Windows protocol checks, installed-wheel qualification, Qwen21 check-only execution, or any model/device experiment. No R3 candidate was supplied/executed here. The original Issue acceptance criteria have not been independently re-authenticated in this increment; the new proposal does not establish their replacement or permission to weaken them. No files were changed and no campaign was resumed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:processRepo-wide process, governance, privacy and security ruleskind:taskBounded delivery task with a delegated slice

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions