@tiangong-ai/cli 0.0.58 → 0.0.60

package/AGENTS.md

@@ -17,8 +17,8 @@ checkPaths:
   - .docpact/config.yaml
   - docs/agents/**
   - src/**
-lastReviewedAt: 2026-09-02
-lastReviewedCommit: 8a18ba69f432d2c639517549157ef29545722cff
+lastReviewedAt: 2026-09-03
+lastReviewedCommit: 06dadef170ab79c19eef9a35663e86fdee22479b
 ---
 
 # Tiangong AI CLI Contract
@@ -126,3 +126,9 @@ test, coverage, and pack checks before publishing.
   the Skills/Research data boundary.
 - Read `docs/agents/data-runtime-implementation-plan.md` before starting or
   sequencing the TypeScript 7 and atomic data migration work packages.
+
+## User Feedback
+
+`CONTRIBUTING.md` implements the shared workspace reporting policy. Keep the
+core form fields aligned with Skills and preserve offline help and packaged
+reporting guidance. User reports do not require a root-cause diagnosis.

package/CONTRIBUTING.md

@@ -0,0 +1,79 @@
+---
+docType: guide
+scope: repo
+status: current
+authoritative: true
+owner: cli
+language: en
+whenToUse: "When reporting bugs or requesting capabilities in cli."
+whenToUpdate: "When reporting forms, feedback entrypoints, or the shared reporting contract change."
+checkPaths:
+  - CONTRIBUTING.md
+  - .github/ISSUE_TEMPLATE/**
+  - README.md
+lastReviewedAt: 2026-09-02
+lastReviewedCommit: 4472738acc78eab396fb1f4c7f5e76af9ac703be
+---
+
+# Contributing Feedback / 提交反馈
+
+Chinese and English reports are welcome. Start with the
+[Bug report](https://github.com/tiangong-ai/cli/issues/new?template=bug_report.yml)
+or [Feature request](https://github.com/tiangong-ai/cli/issues/new?template=feature_request.yml).
+These implement the shared [reporting contract v1](https://github.com/tiangong-ai/workspace/blob/main/_docs/contracts/issue-reporting-policy.md).
+本仓库与另一个项目使用同一套核心字段；可以用中文或英文填写。
+
+## Where To Report / 提交到哪里
+
+- Skill instructions, prompt orchestration, references, and agent routing:
+  [Skills](https://github.com/tiangong-ai/skills/issues/new/choose).
+- Commands, setup/install failures, runtime errors, locks, and package issues:
+  [CLI](https://github.com/tiangong-ai/cli/issues/new/choose).
+- Uncertain or cross-component problems: use CLI and select `Unsure / 不确定`
+  when the component is unknown. Maintainers handle transfers or linked tasks.
+  无法判断归属时提交到 CLI；不需要在两个仓库重复提交。
+
+## Useful Reports / 填写原则
+
+Search existing issues first. Report one problem or capability at a time.
+Describe what you wanted to do, what you expected, and what actually happened.
+A root-cause diagnosis or implementation plan is optional. Unknown versions and
+unreproducible failures are accepted: explain what is known and what is missing.
+先查重复，聚焦一个问题；无需先找出根因，无法复现也可提交已有证据。
+
+For bugs, fill Summary, Goal, Component, Versions, Environment, Stage,
+Reproduction, Expected result, and Actual result. Evidence is optional. Include
+the actual CLI and Skill versions used by the failing run, native host/model,
+OS, and installation method when known. A managed workspace's locked CLI can
+differ from `tiangong-ai --version` in a global shell; use existing run records
+or read-only lock inspection and label the source. Do not update the workspace
+just to obtain a version. 填写出错时使用的实际版本，不要用升级后的版本替代。
+
+For features, fill Summary, Component, Use case, Current limitation, Proposed
+capability, and Success criteria; Alternatives is optional. No fault logs are
+required. 功能建议说明场景、障碍、期望能力和成功示例即可。
+
+Share only minimal sanitized logs, commands, or prompt excerpts. Remove keys,
+tokens, authorization headers, private paths, private research data, and
+unrelated conversation text. Do not attach an entire research directory or
+environment dump. Optional audit exports require review before sharing; do not
+run paid provider/model checks or repeat research solely to file an issue.
+仅提供最少量脱敏证据，不要上传密钥、整个工作区或完整私有对话。
+
+## Agent-Assisted Reports / 让 Agent 协助
+
+Ask the installed Auto Research Skill to prepare an issue report. It carries a
+local reporting reference and Markdown templates; a workspace source checkout
+is unnecessary. Preserve the same English section labels as the form while
+writing the content in either language. State `Unknown` or `Not applicable`
+with a reason instead of inventing details. An agent may draft locally; it posts
+or uploads only with user authorization, including authorization already given
+in the session. 可以先生成草稿；实际提交遵循用户已给出的授权。
+
+## Maintainer Handoff / 维护者接手
+
+Maintainers preserve the original report, verify component ownership, link
+duplicates, and add development scope, TODOs, acceptance criteria, validation,
+and integration requirements before starting implementation. Reporters do not
+need Project access or knowledge of internal delivery rules.
+维护者负责把反馈整理为可执行任务，不把内部交付要求转嫁给反馈者。

package/README.md

@@ -12,14 +12,24 @@ checkPaths:
   - package.json
   - bin/**
   - src/**
-lastReviewedAt: 2026-09-02
-lastReviewedCommit: 8a18ba69f432d2c639517549157ef29545722cff
+lastReviewedAt: 2026-09-03
+lastReviewedCommit: 06dadef170ab79c19eef9a35663e86fdee22479b
 ---
 
 # Tiangong AI CLI
 
 Package: `@tiangong-ai/cli` Executable: `tiangong-ai` Node: `>=24`
 
+## Feedback
+
+Use the [Bug and feature forms](https://github.com/tiangong-ai/cli/issues/new/choose)
+and [reporting guide](https://github.com/tiangong-ai/cli/blob/main/CONTRIBUTING.md).
+Chinese and English reports are welcome; unknown versions or incomplete
+reproduction are accepted with an explanation. CLI help exposes these links,
+and the npm package includes `CONTRIBUTING.md`. For Skill instructions and
+orchestration, use the [Skills forms](https://github.com/tiangong-ai/skills/issues/new/choose);
+uncertain ownership can be reported here for maintainer triage.
+
 ## Run From This Repository
 
 ```bash
@@ -702,9 +712,15 @@ submission manifest must bind distinct absolute files for cover letter, title
 page, reporting checklist, data availability, code availability, and source
 data; figure/table index, extended data, and supplementary methods are optional.
 `research publication freeze` then content-addresses the Policy, scientific
-design and early reviews, acquisition/content/inference snapshots, reproduced
+design and early reviews, acquisition/content/inference snapshots, mode-bound
 analysis, Claim-Evidence Graph, base outputs, manuscript, assessment,
 supplements, role-complete submission files, and reproducibility manifest.
+Computational/mixed analysis still requires reproduced metadata with exact
+implementation/environment bindings. Qualitative analysis uses
+`status: not-applicable`, null command/seed and empty implementation/environment
+lists; it must not invent a computation. Both paths retain the same evidence,
+graph, Policy and independent-review checks. Metadata alone is not proof that
+a computation was executed.
 Exactly four fresh independent sessions review that frozen generation:
 evidence, methods/reproducibility, domain/novelty, and journal-editor. A revised
 manuscript invalidates prior reviews. Every reviewer must use the configured
@@ -1006,6 +1022,14 @@ deficits; exit `0` means only potential eligibility, never successful atom
 registration, content freeze, independence certification, or review. Pending
 input materialization, decomposition, and exact atom assignments remain
 explicit. Re-run after material acquisition changes, not after every atom.
+`submissionGate` separately reports deterministic acquire-submit blockers;
+`potentially-ready` is not an inference/content pass. An accepted binary without
+required readable content stops before submission, while an honest incomplete
+audit may still freeze its limitations and separate inference-stop decision.
+An admitted local PDF/Office file alone is not producer-readable. Forecast uses
+the same media rules as artifact registration and reports a missing readable
+derivative unless a verified readable artifact or admitted text/context input
+is available. This classification does not reread or decode the binary again.
 Flat `sourceTypeRequirements` arrays mean **all-of**. A design may instead use
 `{"allOf":["academic-paper"],"anyOf":["government","industry"],"atLeast":{"count":2,"from":["academic-paper","government","industry"]}}`;
 every present group applies, and counts use distinct types. Forecast, typed
@@ -1360,27 +1384,118 @@ read-only under `outputs/revisions/synthesize/<sha256>/report.md`; review and
 closure are invalidated, while unchanged discovery, acquisition, content,
 inference, analysis, and graph objects remain bound.
 
-Recovery fork writes are rollback-protected. If inherited output, content,
-inference, or graph validation fails, the target directory is removed and
-source authority is restored. A source typed-content snapshot is revalidated
+Recovery forks have one journal commit point. Before `project.forked` commits,
+their target is staging, not an authoritative project; the original source
+remains authoritative. Status, run and native/reviewer admission share that
+same verified lineage view. A source typed-content snapshot is revalidated
 and re-signed for the target generation instead of copying its project-bound
 hash. Top-journal generations may resume only through `acquire`, because the
 target-specific scientific design must complete fresh evidence-construct and
-pilot-methods reviews before analysis. `research status --all` marks any legacy
-fork directory without a `project.forked` commit marker as
-`authority.state = "invalid"` rather than authoritative.
-
-To repair completed acquisition before analysis, use
-`research project fork SOURCE --to TARGET --resume-through discover`. This
-preserves the original frozen audit, inherits verified discovery/receipts and
-exact acquired artifacts, and opens TARGET's acquire stage. Reuse the returned
-artifact IDs from the source acquisition audit, run
-`project evidence content forecast TARGET --input AUDIT`,
-then prepare/submit TARGET's acquire stage and rebuild typed content. There is
-no need to repeat paid searches or download unchanged files. Top-journal
-recovery additionally requires a Policy approved for TARGET and `--design`,
-`--design-producer-agent`, and `--design-producer-session`; new generation
-reviews cannot inherit the source generation's scientific approval.
+pilot-methods reviews before analysis.
+
+After process interruption, repeat the same explicit fork request. The next
+lease holder settles only hash-bound CLI-owned pending operations: before
+commit it retains the interrupted target under
+`lineage/interrupted-project-mutations/<operation-id>/target`, outside the
+project namespace; after commit it finishes source/ledger projections. Exact
+committed replay returns the existing target without recopying evidence or
+repeating provider work. A different request for that target is a conflict.
+An unchanged completed retry request is likewise acknowledged without another
+revision or attempt. A subsequent real failure can still be retried normally.
+
+Unknown targets, symlink replacements, modified source state and invalid
+recovery metadata are not overwritten or deleted. Resolve the reported
+`RESEARCH_PROJECT_RECOVERY_REQUIRED` conflict or restore a trusted backup before
+retrying; do not forge control files. Read-only status ignores never-committed
+directories without project state and reports existing uncommitted derived
+states as invalid. Missing state for a committed project is still an error.
+Recovery does not truncate a corrupt journal or claim power-loss durability.
+
+To repair completed acquisition before any analysis attempt, prefer an explicit
+same-project revision when question, Policy, design, and requirements are unchanged:
+
+```bash
+tiangong-ai research project evidence acquisition revise PROJECT \
+  --expected-snapshot <current-acquisition-sha256> \
+  --reason "Add the verified readable derivative before analysis" \
+  --workspace /absolute/workspace --json
+```
+
+This reopens acquire, preserves discovery, exact artifacts/receipts, historical
+snapshot bytes, budgets already spent, and the research-design approval. It
+invalidates evidence-construct and pilot-methods approvals because their evidence
+changed. Add `--include-discovery` only when new sources must be discovered or
+admitted; that explicitly reopens discover then acquire without a new project or
+automatic provider call. Prepare the returned stage, reuse unchanged objects,
+forecast once per meaningful batch, submit the complete revised audit, and rebuild
+typed content. Exact revision replay is idempotent. Stale snapshots, active native
+sessions, unresolved handoffs, inference freezes, and later attempts are refused.
+
+Changed failed/limited decomposition records can be superseded under a descendant
+acquisition snapshot; unchanged records are reused. Single and batch intake share
+this rule. Historical records remain immutable, and atoms from deselected artifacts
+cannot fill current coverage. Same-snapshot conflicting extraction still fails.
+
+Actual question/Policy/design changes or post-analysis work require the existing
+fork/addendum flow. Pre-feature snapshots without immutable evidence records cannot
+be repaired in place; use `research project fork SOURCE --to TARGET
+--resume-through discover` to reuse discovery/receipts/artifacts, or explicitly
+start a new generation. There is no automatic migration. A top-journal successor
+requires a Policy approved for TARGET and `--design`, `--design-producer-agent`,
+and `--design-producer-session`; it cannot inherit scientific approval.
+
+### Original task, current scope, and actual checks
+
+For a new research project, record a small original-requirement checklist after
+project initialization and before any producer execution or scientific review.
+The native host authors the content; the CLI owns its schema and hash bindings:
+
+```bash
+tiangong-ai research schema show task-contract --json
+tiangong-ai research project task define PROJECT \
+  --input /absolute/task.json --workspace /absolute/workspace --json
+tiangong-ai research project task status PROJECT --workspace /absolute/workspace --json
+```
+
+Each requirement has a stable ID, acceptance condition, `checkKind` (`evidence`,
+`computation`, or `proof`), and optional bindings to existing design claims and
+coverage dimensions. Original wording cannot be overwritten. Old projects without
+a task remain explicitly unassessed rather than retrospectively accepted.
+
+Before analysis, use `research schema show task-scope-change` and
+`project task scope propose PROJECT --input FILE --expected-contract SHA` to
+propose a change. Review the returned `changes.details` before/after values, then
+approve only with both `--proposal SHA` and `--confirm-change SHA` on
+`project task scope approve`. A generic continue instruction is not scope consent.
+The record is an **operator confirmation**, not authenticated proof of a human's
+identity. Scope approval does not rewrite scientific design or evidence floors;
+it invalidates prior scoped scientific reviews. Withdrawn original requirements
+remain visible and a fork/addendum retains original history without inheriting
+task completion.
+
+Use `research schema show task-acceptance` and
+`project task acceptance record PROJECT --input FILE` after acquisition and
+between native stages. Records bind the exact requirement version, source/atom/
+finding IDs, and explicitly selected bounded UTF-8 result files. The declared
+command is stored only by hash and is **not executed by this command**. Raw result
+bytes are copied into immutable hash-addressed objects; secrets and control-store
+sources are rejected. Positive/negative computational outcomes need a command and
+results; failed, inconclusive, and not-run checks remain honest without invented
+results. All records say `trust=native-observation`, `executionCertified=false`.
+
+One unchanged result blob is stored and included once per reviewer context. The
+existing independent review receives the original request, original/current
+requirements, exact checks and results, and returns a bound `taskAssessment`;
+there is no additional default paid review round. Missing current checks stop
+before review, and stale/failed/inconclusive checks cannot be promoted to answered.
+Publication packets and portable audit verification retain these relationships;
+hash integrity does not prove execution, scientific validity, or editorial acceptance.
+
+`project task status` and each `research run` project summary report original and
+current task completion separately from workflow completion and publication verdict.
+A completed workflow or approved reduced scope must not be described as satisfying
+unanswered original requirements. Use the selected runtime's help/schema discovery
+once before adopting these commands; they are not an implicit runtime upgrade.
 
 ## Research Search
 

package/dist/cli.js

@@ -2733,6 +2733,11 @@ Run "tiangong-ai kb --help" for KB options.
 Run "tiangong-ai data --help" for atomic data options.
 Run "tiangong-ai research --help" for research options.
 Run "tiangong-ai education --help" for education options.
+
+Report a problem or suggest a capability:
+  https://github.com/tiangong-ai/cli/issues/new/choose
+Reporting guide:
+  https://github.com/tiangong-ai/cli/blob/main/CONTRIBUTING.md
 `;
 }
 function kbHelp() {