feat(walkthrough): add agent onboarding skill-delivery walkthrough sm… - #3131
feat(walkthrough): add agent onboarding skill-delivery walkthrough sm…#3131rootkiller6788 wants to merge 1 commit into
Conversation
…oke (GH-C79) Coverage of canonical release and active project-skill manifests: - Packaged skill catalog (5 packaged + 1 generated = 6 required, PR program skill present) - Skill install readback lifecycle (build, write, inspect on disk) - Missing skill detection (3 of 5 packaged skills correctly missing) - Stale skill detection (modified content vs readback mismatch) - Install dedupe (retire_duplicate_managed_skills from alternate root) - CWD isolation (LOOPX_SKILLS_DIR from env, not cwd) - Skill delivery mode parity across 13 agent types (3 host_managed, 10 surface_managed) - Onboarding packet includes skill delivery contract per agent type - Readback integrity states (null dir, missing manifest, stale content) - Public safety across all outputs
huangruiteng
left a comment
huangruiteng
left a comment
There was a problem hiding this comment.
详细中文评审（durable_smoke_value 门槛）
精确评审头： 3131@bcb00cf1b62ef2478fd7f8caab9f9e8ad0b7ca9a
动机
本 PR 是单文件 examples/agent-onboarding-skill-delivery-walkthrough-smoke.py（+585 行），演示 agent onboarding 的 skill 安装/readback/stale 检测/去重/CWD 隔离。按新的 durable_smoke_value 门槛，例程类 PR 必须证明对仓库/产品的真实、持久价值；能运行、确定性、public-safe 只是必要条件。本 PR 不满足该门槛。
改动思路
10 个 scenario 覆盖打包 skill 目录、install readback、missing/stale 检测、去重、CWD 隔离、13 种 agent 类型的 delivery mode 与 public safety。作为演示自洽，但全部是对现有 loopx.skill_install_readback 与 onboarding 行为的合成断言，没有新增产品行为，也没有把断言收敛为 focused test。
具体改动（关键内容讲解）
- existing_coverage_scan：
tests/test_ark_managed_agent_host.py已覆盖skill_install_readback生命周期；tests/test_agent_onboarding_unconnected_project.py、tests/test_host_loop_activation.py、tests/test_gemini_cursor_host_surfaces.py与examples/control_plane/agent-onboard-host-loop-activation-smoke.py已覆盖 onboarding/skill delivery 的关键路径。本 walkthrough 的 10 个 scenario 与这些既有覆盖高度重叠，属于重复演示而非差异化覆盖。 - same_author_batch_scan：#3130（805 行）、#3131（585 行）、#3133（545 行）由同一作者在 11:18–11:20 两分钟内连续提交，且该作者此前已有 #3113–#3120 一整串同构 walkthrough PR。这是明确的批量刷 PR 信号。
- real_product_or_repo_value_verdict：若 readback/stale/去重语义确实缺少 thin 断言，正确做法是在
tests/里加 focused pytest；用 585 行 example 脚手架重复既有覆盖，对仓库是噪音。
对主干的风险
代码本身无运行时风险（example-only、public-safe、可运行）。风险在模式：同作者以分钟级频率堆叠同形状 walkthrough PR，消耗评审队列、抬高维护成本，并让真实 runtime 改动被淹没。照此模式继续，队列中的同类 PR 会被模板化放行，这正是本 gate 要拦截的。
我的整体评价
REQUEST_CHANGES。 修复建议（择一）：(1) 将本 smoke 收敛为 thin focused pytest（tests/test_skill_install_readback.py 风格），只保留 readback/stale/去重/CWD 隔离的关键断言；(2) 与 #3130/#3133 合并为一个有差异化价值的 walkthrough，删除重复 scaffolding。贡献行为警告：按 PR #3134 的新政策，若同一作者在收到本警告后再提交此类低价值同形状 PR，将禁止该账号继续向本仓库提交 PR（owner 执行屏蔽）。
English Verdict
REQUEST_CHANGES — exact head bcb00cf1b62ef2478fd7f8caab9f9e8ad0b7ca9a.
Under the durable_smoke_value gate, this 585-line example-only walkthrough does not prove real, durable value: the skill-install readback/stale/dedupe/CWD behavior is already covered by existing tests (test_ark_managed_agent_host.py and onboarding smokes), and it is part of a same-author 3-PR batch opened within two minutes (#3130/#3131/#3133) on top of the #3113–#3120 walkthrough series. Consolidate into thin focused tests or one differentiated PR. Contribution warning: further low-value same-shape PRs after this warning will lead to the account being blocked from submitting PRs to this repository.
2 participants
…oke (GH-C79)
Coverage of canonical release and active project-skill manifests:
Summary
Issue Or Task
Validation
python3 -m py_compile loopx/*.pyloopx check --scan-root .Type of Change
LoopX Area
Boundary Checklist
.loopx/,.codex/goals/, liveACTIVE_GOAL_STATE.md, credentials, private benchmark traces, verifier output, raw agent sessions, internal document links, or local machine paths.