Harness Agent · 一个 manager,模型判断意义,确定性系统保证执行真相
TIER C · CONTRACT ONE MANAGER ONE PRIMARY MODE AI-NATIVE
TIER C · CONTRACT ONE MANAGER ONE PRIMARY MODE AI-NATIVE
NovelForge Harness 把经过验证的小说 Project 与明确任务组织成一轮 bounded、resumable run。它拥有 execution policy,不拥有 story truth。
Project authority 拥有 Canon 与项目专属事实;模型拥有 semantic fiction judgment;deterministic runtime 拥有 identity、power、persistence 与 exact execution state。
01 · AI-native 不等于 model-authoritative
Section titled “01 · AI-native 不等于 model-authoritative”需要理解意义的工作默认由模型承担,例如:
- search intent、query formulation、relevance、continuation / stopping;
- story / scene / planning interpretation;
- character action、motivation、plausible inference、integrity;
- Reader experience;
- semantic hard-rule applicability / violation;
- repair mechanism / depth;
- research interpretation;
- feedback / preference interpretation。
Deterministic code 只保留可机械证明的不变量:
- Project/resource/session/run/checkpoint identity;
- permissions / capability boundaries;
- exact artifacts、hashes、fingerprints;
- provenance / exact-source binding;
- stage / private-state visibility;
- persistence、CAS、transaction、idempotency;
- hard budgets / resource limits;
- typed envelope/result/receipt validation;
- required semantic execution 是否真实发生,以及是否绑定 exact candidate;
- settlement / release-role invariant。
Semantic result 默认只是 evidence / proposal,除非另外的 authority mechanism 明确授予更多权力。
02 · 默认只用一个 manager
Section titled “02 · 默认只用一个 manager”优先一个能力足够强的 manager/agent。只有以下情况能证明收益时才拆分:
- mandatory independent evaluation;
- information/context isolation;
- per-character private state;
- genuinely different permission/tool/runtime;
- immutable input 上真正有价值的 parallel work;
- human review。
不要为了模仿软件组织架构而制造 multi-agent round-table。
03 · 恰好一个 primary task mode
Section titled “03 · 恰好一个 primary task mode”每个 user-visible run 只能有一个 primary mode:
DESIGN-BOOK | DESIGN-VOLUME | PLAN-UNIT | PLAN-CHAPTER | DRAFT | REVISE | RESEARCH | SETTLE | AUDIT | CORPUS-INGEST | LEARN | SYSTEM-IMPROVE
Mode 可以内部调用共享 subroutine,但不能悄悄执行另一个 mode 的 user-visible side effect。DRAFT 不自动 SETTLE;AUDIT 不偷改 manuscript;LEARN 不自授 durable behavior。
04 · Semantic work 前先 bootstrap live authority
Section titled “04 · Semantic work 前先 bootstrap live authority”Fresh manager 依次解析:
- current/pinned Framework manifest / identity;
- consuming Project manifest + exact lock/fingerprint;
- Project Adapter / logical paths;
- 只能有一个 task mode;
- manager session/run identity;
- authority cutoff + permissions;
- sparse Context Manifest / candidate set;
- current host capabilities。
旧聊天 / provider session 不能替代 bootstrap authority。Resume 必须重新验证 Framework/Project compatibility、current fingerprints、approval/write precondition 与 pending capabilities。
05 · Search / Context:语义选择,确定性边界
Section titled “05 · Search / Context:语义选择,确定性边界”Manager/model 自己判断缺什么、搜什么、query 怎么写、哪个结果真正相关、是否 reformulate / continue、什么保留、什么时候 evidence 足够。
Runtime 可以提供 search/fetch/extract/index primitive 与 candidate generation,但 recency、fixed last-N、vector similarity、top-k、item class 都不能冒充 narrative truth。
context_inspector.py 只检查 mechanical eligibility / stage / protected-edit state,并明确拒绝 relevance 字段。
context_assembly.py v2 在 semantic selection 之后运行,只验证 exact selected/source refs、receiving stage、hidden/private/invalidated state、fingerprints,以及操作在机械意义上确实要求时的 exact higher-authority refs。它不再执行 literary class/purpose obligation,也不宣称 semantic sufficiency。
Hard-budget packing 可以在 selection 后执行 whole-item resource limit。Persistent storage 不代表 automatic prompt injection。
06 · DRAFT / REVISE production responsibilities
Section titled “06 · DRAFT / REVISE production responsibilities”默认 adaptive path 大致是:
authority/session bootstrap→ agent-owned context/search→ deterministic exact-set/stage/fingerprint verification→ Story / Canon + planning preflight→ private character state→ character.action_propose→ scene.resolve_actions→ compact writer-safe realization→ Reader Pressure→ event-first Raw Draft→ Surface realization→ freeze candidate fingerprint→ Blind Reader (`reader.engagement_audit`)→ Semantic Rule Auditor when required (`quality.semantic_rule_audit`)→ Editor repair spec (`editor.repair_spec`)→ repair / challenger comparison as warranted→ continuity/state checks→ required independent semantic gate→ user-visible Review Draft必须保留这些边界:
- Raw Draft 是内部产物;
- regression bad examples 在 Raw Draft/candidate freeze 前不得进入 Writer context;
- private character state 是 causal evidence,不是 Writer exposition payload;
- Surface clean 只是地板,不是 production readiness;
- Blind Reader 不是 hard-rule checklist executor;
- Rule Auditor 获得 Reader 不该看到的 authoritative rule material;
- Editor 语义上选择 repair owner 与 generation mode;
repair_policy.py只执行所选 mode 对 writer-context 的信息边界;- material candidate change 会使旧 fingerprint-bound review result 失效;
- explicit acceptance 与 SETTLE 继续分离。
07 · Model-readable semantic contracts
Section titled “07 · Model-readable semantic contracts”Catalog authority:harness/semantic_workers/model_contract_catalog.json。
Semantic work 通过这个 progressive-disclosure catalog 精确解析 contract ID。Job 绑定:
- kind / subject / exact contract version;
- bounded input/context;
- rubric;
- output contract;
- permissions;
- semantic fingerprint;
- execution provenance requirement。
模型负责 judgment。Runtime 只验证 identity/fingerprint/permission/type/provenance,并 consume-once。内部 semantic work 不自动获得 independent 属性。
08 · Capability broker
Section titled “08 · Capability broker”任何 tool / external action 都必须针对 current host capabilities 重新解析。Undeclared capability 视为 unavailable。
Provider name、PATH 上的 executable、旧 session 记忆、network primitive、documentation page、model self-assertion 都不是 authorization proof。
Capability 也不等于 authority:filesystem write capability 不代表 Canon write permission。
Credential / authority token 不进入普通 semantic context。
09 · Durable session != model context
Section titled “09 · Durable session != model context”Execution identity 保持分层:
Project/resource→ session→ run→ checkpoint→ event/handoff/job→ result→ validated consume-once receipt→ resumeExternal wait、required independent review、consequential write 前按合同 checkpoint。Context-window 丢失可以恢复,因为 authority state 存在 durable artifacts/events/checkpoints,而不是聊天 transcript。
10 · Independent semantic integrity
Section titled “10 · Independent semantic integrity”Gate 真正要求 independence 时,manager 可以:
freeze → package → checkpoint → dispatch → await → validate → consume → route repair
但不能换个 internal role label 就自己完成 judgment。
Materially changed candidate 默认需要新的 bound review,除非合同显式允许 reuse。Transport failure 可切换 eligible transport;有效的语义拒绝必须进入 repair,也不能反复更换评审直到有人接受 candidate。
没有 eligible independent provider/model 时必须 PENDING_MODEL,绝不能 PASS。
11 · LEARN / CORPUS / SYSTEM-IMPROVE
Section titled “11 · LEARN / CORPUS / SYSTEM-IMPROVE”Learning 必须分开:
semantic interpretation!= evidence!= hypothesis!= promotion review!= write authority!= active eligibility!= current relevancelearning.preference_interpret 解释 feedback;LearningStore / Author Model 在 authority/CAS 下持久化 evidence/hypothesis。Promotion Gate 验证 exact bound semantic review 与客观 prerequisite,不用 arbitrary evidence-count threshold 冒充 semantic truth,也不能授予 write authority。
Active preference 不会自动注入;manager/model 为当前任务显式选择 relevant active hypothesis IDs。
General Craft change 继续属于 SYSTEM-IMPROVE,必须有 current research、counterexamples/evals、compatibility、rollback 与 explicit promotion authority。
12 · Writes / Settlement
Section titled “12 · Writes / Settlement”每个 consequential side effect 都需要 least privilege、exact target、before-state/precondition、idempotency、必要时 checkpoint/write intent、postcondition verification、trace/rollback semantics。
Canon settlement 只有在 explicit Project acceptance / authorized Canon intent 后才合法。Connector、schedule、webhook、model result、learning state、CI、corpus evidence 都不会因为“到达系统”就自动授予 write authority。
13 · Truthful states
Section titled “13 · Truthful states”合法状态包括:
complete · review · awaiting_user · awaiting_external · semantic_pending · semantic_invalid · failed_gate · blocked · settlement_incomplete
Green deterministic workflow 不能替代 required semantic judgment;没有真正执行的 semantic job 继续 pending。Required gate 未解决时不得声称 production-ready。
14 · SYSTEM-IMPROVE execution discipline
Section titled “14 · SYSTEM-IMPROVE execution discipline”Material Framework work 按以下链路执行:
live bootstrap→ candidate/owner reconciliation→ current research→ overreach audit→ spec/plan/tasks→ incremental implementation→ ablations + deterministic tests→ independent semantic evidence when required→ CI/security/compatibility/docs synchronization→ human-review readiness每次 consequential write 前重新验证 current branch/HEAD/before-state。Long operation 必须 bounded;同一 pending 状态没有新 evidence 时转入 jobs/logs diagnosis,而不是 blind waiting。Candidate-owned failure 与 pre-existing unrelated repository debt 必须区分。