Conradgui · Aug 2, 2026 · Aug 2, 2026 · Aug 2, 2026 · Aug 2, 2026 · Aug 2, 2026
diff --git a/.github/ISSUE_TEMPLATE/bug_report.yml b/.github/ISSUE_TEMPLATE/bug_report.yml
@@ -0,0 +1,76 @@
+name: Bug report
+description: Report a reproducible Loop Craft defect without exposing private source material.
+title: "[Bug]: "
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for helping improve Loop Craft. For suspected security issues, stop here and use the private reporting link shown in the template chooser.
+  - type: dropdown
+    id: version
+    attributes:
+      label: Loop Craft version
+      options:
+        - v0.4.0
+        - main
+        - Other commit or version
+    validations:
+      required: true
+  - type: dropdown
+    id: route
+    attributes:
+      label: Entry route
+      options:
+        - From scratch
+        - Existing Skill
+        - Conversation
+        - Direct Build
+        - Verify
+    validations:
+      required: true
+  - type: dropdown
+    id: adapter
+    attributes:
+      label: Selected Adapter
+      options:
+        - codex-skill
+        - compact-prompt
+        - Not reached or not applicable
+    validations:
+      required: true
+  - type: textarea
+    id: reproduction
+    attributes:
+      label: Minimal reproduction
+      description: Use synthetic inputs and include the exact command or Agent request.
+      placeholder: Describe the smallest sequence that reproduces the problem.
+    validations:
+      required: true
+  - type: textarea
+    id: expected
+    attributes:
+      label: Expected behavior
+    validations:
+      required: true
+  - type: textarea
+    id: actual
+    attributes:
+      label: Actual behavior
+    validations:
+      required: true
+  - type: textarea
+    id: environment
+    attributes:
+      label: Environment
+      description: Operating system, Python version and Agent surface.
+    validations:
+      required: true
+  - type: checkboxes
+    id: redaction
+    attributes:
+      label: Evidence redaction
+      options:
+        - label: I removed secrets, personal data, raw private conversations, absolute local paths and unrelated Evidence.
+          required: true
+        - label: This is not a security vulnerability or an authority-boundary bypass.
+          required: true
diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml
@@ -0,0 +1,8 @@
+blank_issues_enabled: false
+contact_links:
+  - name: Private security report
+    url: https://github.com/Conradgui/loop-craft/security/advisories/new
+    about: Report vulnerabilities and authority-boundary bypasses without exposing sensitive material.
+  - name: Usage and product boundaries
+    url: https://github.com/Conradgui/loop-craft#readme
+    about: Check installation, routes, current capabilities and explicit limitations first.
diff --git a/.github/ISSUE_TEMPLATE/feature_request.yml b/.github/ISSUE_TEMPLATE/feature_request.yml
@@ -0,0 +1,50 @@
+name: Feature request
+description: Propose a bounded user capability, not infrastructure without a user path.
+title: "[Feature]: "
+body:
+  - type: textarea
+    id: user_problem
+    attributes:
+      label: User problem and value
+      description: What can a user not complete today, and who encounters the problem?
+    validations:
+      required: true
+  - type: textarea
+    id: current_alternative
+    attributes:
+      label: Current alternative
+      description: How is the task handled today, and why is that insufficient?
+    validations:
+      required: true
+  - type: dropdown
+    id: boundary
+    attributes:
+      label: Product boundary
+      options:
+        - Existing 0/1-Loop and Adapter behavior
+        - New entry or packaging behavior
+        - Runtime, multi-Loop, publishing or another currently unsupported area
+        - Unsure
+    validations:
+      required: true
+  - type: textarea
+    id: proposal
+    attributes:
+      label: Proposed outcome
+      description: Describe observable user behavior, not a preferred internal implementation.
+    validations:
+      required: true
+  - type: textarea
+    id: evidence
+    attributes:
+      label: Evidence and examples
+      description: Provide sanitized examples, repeated failure patterns or prior attempts.
+  - type: checkboxes
+    id: scope
+    attributes:
+      label: Scope check
+      options:
+        - label: I checked the README boundaries and existing Issues before proposing this capability.
+          required: true
+        - label: I removed private conversations, credentials, personal data and unrelated Evidence.
+          required: true
diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
@@ -0,0 +1,25 @@
+## User impact
+
+What can a user do after this change that they could not do before?
+
+## Scope
+
+- Entry, Core, Adapter, Evidence or repository surface changed:
+- Source and authority boundaries affected:
+- Deliberately unchanged:
+
+## Validation
+
+- [ ] Targeted deterministic checks:
+- [ ] Real build / verify when the build chain changed:
+- [ ] Behavioral evaluation when shared Skill prose changed:
+- [ ] Local paths, conversations, credentials and unrelated Evidence are absent from the diff:
+
+## Not run
+
+List checks deliberately not run and explain which concrete risk did not justify them.
+
+## Boundaries
+
+Document unsupported behavior, compatibility limitations and any follow-up that is intentionally
+outside this Pull Request.
diff --git a/.github/workflows/validate.yml b/.github/workflows/validate.yml
@@ -2,7 +2,12 @@ name: Validate Loop Craft
 
 on:
   push:
+    branches: [main]
   pull_request:
+  workflow_dispatch:
+
+permissions:
+  contents: read
 
 jobs:
   validate:
@@ -20,14 +25,15 @@ jobs:
       PYTHONUTF8: "1"
 
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
 
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
         with:
           repository: Conradgui/skill-creator-pro
+          ref: eb23656e56ea3555599a6c5278a8b5834dc56b6d
           path: .tmp/skill-creator-pro
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
         with:
           python-version: ${{ matrix.python-version }}
 

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -18,6 +18,8 @@ Dual-output Adapter release.
   normalized result digests.
 - Adapter-neutral Artifact contracts and adapter-aware read-only verification for both Skill
   and Prompt drift.
+- A stable `v0.4.0/loop-craft` installation path plus Security, Code of Conduct, structured
+  Issue Forms and a Pull Request contract for public collaboration.
 
 ### Changed
 
@@ -28,14 +30,16 @@ Dual-output Adapter release.
 - Final staging promotion has a bounded retry for the observed transient Windows
   `PermissionError`; persistent permission failures still stop the build.
 - Version metadata is unified at `0.4.0`.
+- GitHub Actions and the Skill Creator PRO validator source are pinned to immutable revisions;
+  the workflow can also be invoked manually as a release gate.
 
 ### Boundaries
 
 - Compact Prompt preserves the approved behavior, authority, verification, stopping and
   invariant content, but it does not supply tools, state, scheduling, or independent audit.
 - Source-preserving Existing Skill upgrades remain Codex Skill builds. Runtime, multi-Loop,
-  Override, Subloop, publishing, scheduling, installation automation, Release and Git tags
-  remain outside this version.
+  Override, Subloop, publishing, scheduling and installation automation remain outside this
+  version.
 
 ## 0.3.0 - 2026-07-31
 

diff --git a/CODE_OF_CONDUCT.md b/CODE_OF_CONDUCT.md
@@ -0,0 +1,29 @@
+# Code of Conduct
+
+## Our standard
+
+Loop Craft welcomes people who participate in good faith, regardless of experience level or
+background. Be respectful, focus criticism on ideas and observable evidence, explain unfamiliar
+terms, and leave room for people to correct mistakes.
+
+Unacceptable behavior includes harassment, discrimination, personal attacks, threats, deliberate
+misrepresentation of another person's work, publishing private information without permission,
+or repeatedly disrupting a technical discussion after a maintainer has set a boundary.
+
+## Project scope
+
+This standard applies in the repository and in public spaces where someone is representing Loop
+Craft. Technical disagreement is welcome. Maintainers may edit, remove or reject contributions
+and may temporarily or permanently restrict participation when behavior is unsafe or persistently
+disruptive.
+
+## Reporting and enforcement
+
+For an ordinary moderation concern, flag the relevant GitHub content or open a concise Issue. If
+the report contains private or sensitive information, use the
+[private reporting form](https://github.com/Conradgui/loop-craft/security/advisories/new) and begin
+the title with `CODE OF CONDUCT`.
+
+Reports will be reviewed as privately as the selected GitHub channel permits. The maintainer will
+consider context, impact, prior behavior and the safety of affected participants before choosing
+a proportionate response. Retaliation against a good-faith reporter is not acceptable.
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -127,3 +127,13 @@ the official Skill validator, link checking, a real build/verify, and a drift ne
 on Linux and Windows across Python 3.12 and 3.13.
 
 CI does not cover behavioral evaluation. If you touched prose, say so and attach results.
+
+## Security reports
+
+Do not open a public Issue for a suspected vulnerability, authority-boundary bypass, sensitive
+Evidence leak or approval/stop-condition bypass. Follow [SECURITY.md](SECURITY.md) and use the
+private GitHub reporting form. Public Bug reports must use synthetic or redacted inputs and must
+not contain credentials, personal data, raw private conversations or unrelated local paths.
+
+Participation in Issues and Pull Requests is also governed by the
+[Code of Conduct](CODE_OF_CONDUCT.md).
diff --git a/README.md b/README.md
@@ -17,7 +17,8 @@
 > Workflow instead of manufacturing a fake Loop.
 
 [中文版](README.zh.md) · [Design](docs/DESIGN.md) ·
-[Evaluation](docs/REAL_WORLD_EVALUATION.md) · [Contributing](CONTRIBUTING.md)
+[Evaluation](docs/REAL_WORLD_EVALUATION.md) · [Contributing](CONTRIBUTING.md) ·
+[Security](SECURITY.md)
 
 ---
 
@@ -97,7 +98,7 @@ Ask Codex to use its Skill Installer:
 
 ```text
 Use $skill-installer to install the skill at
-https://github.com/Conradgui/loop-craft/tree/main/loop-craft
+https://github.com/Conradgui/loop-craft/tree/v0.4.0/loop-craft
 ```
 
 Start a new Codex turn after installation so the Skill can be discovered.
@@ -292,6 +293,14 @@ The exact source revisions, adopted mechanisms, exclusions, and ownership bounda
 recorded in the [resource registry](docs/references/resource-registry.yaml) and summarized in
 [NOTICE](NOTICE.md).
 
+## Community
+
+Report reproducible defects through [Issues](https://github.com/Conradgui/loop-craft/issues)
+and propose changes through [Pull Requests](https://github.com/Conradgui/loop-craft/pulls).
+Read [Contributing](CONTRIBUTING.md) and the [Code of Conduct](CODE_OF_CONDUCT.md) before
+participating. Suspected vulnerabilities and authority-boundary bypasses belong in the private
+channel described by the [Security Policy](SECURITY.md), never in a public Issue.
+
 ## License
 
 Loop Craft is distributed under the [Apache License 2.0](LICENSE). Independent reference

diff --git a/README.zh.md b/README.zh.md
@@ -16,7 +16,8 @@
 > “有 Loop”强行制造循环。
 
 [English](README.md) · [架构设计](docs/DESIGN.md) ·
-[评估证据](docs/REAL_WORLD_EVALUATION.md) · [参与开发](CONTRIBUTING.md)
+[评估证据](docs/REAL_WORLD_EVALUATION.md) · [参与开发](CONTRIBUTING.md) ·
+[安全策略](SECURITY.md)
 
 ---
 
@@ -94,7 +95,7 @@ flowchart TB
 
 ```text
 使用 $skill-installer 安装这个地址中的 Skill：
-https://github.com/Conradgui/loop-craft/tree/main/loop-craft
+https://github.com/Conradgui/loop-craft/tree/v0.4.0/loop-craft
 ```
 
 安装后开启一个新的 Codex 对话，让新 Skill 被重新发现。
@@ -266,6 +267,13 @@ Loop Craft 选择性地本地化了 Loopy、Workflow Skill Creator、Skill Polis
 每个来源的固定 revision、采纳机制、排除内容和所有权边界记录在
 [资源登记](docs/references/resource-registry.yaml)，摘要见 [NOTICE](NOTICE.md)。
 
+## 社区参与
+
+可复现缺陷请提交到 [Issues](https://github.com/Conradgui/loop-craft/issues)，代码与文档改动请走
+[Pull Requests](https://github.com/Conradgui/loop-craft/pulls)。参与前请阅读[贡献指南](CONTRIBUTING.md)
+和[行为准则](CODE_OF_CONDUCT.md)。疑似漏洞或权限边界绕过必须使用[安全策略](SECURITY.md)中的
+私密入口，不要公开到 Issue。
+
 ## 许可证
 
 Loop Craft 按 [Apache License 2.0](LICENSE) 分发。独立参考项目继续适用各自的许可证与归属

diff --git a/SECURITY.md b/SECURITY.md
@@ -0,0 +1,41 @@
+# Security Policy
+
+## Supported versions
+
+Security fixes are considered for the current `0.4.x` line. Older development snapshots are
+not supported. The repository may document later versions before a formal Release exists; the
+[Releases page](https://github.com/Conradgui/loop-craft/releases) is the authoritative list of
+published versions.
+
+## Report a vulnerability privately
+
+Do not open a public Issue for a suspected vulnerability. Use GitHub's
+[private vulnerability reporting form](https://github.com/Conradgui/loop-craft/security/advisories/new)
+and include:
+
+- the affected Loop Craft version or commit;
+- the entry route and selected Adapter;
+- the smallest safe reproduction;
+- the expected and observed authority or data boundary;
+- whether generated Artifact or Evidence content is involved;
+- a suggested remediation, when available.
+
+Remove secrets, raw private conversations, credentials, personal data, absolute local paths and
+unrelated Evidence before submitting. A synthetic reproduction is preferred.
+
+The maintainer will make a best-effort acknowledgement within seven days, assess severity and
+coordinate disclosure through the private advisory. Remediation timing depends on impact and
+available maintainer capacity; this is not a guaranteed service-level agreement.
+
+## Security boundary
+
+Loop Craft is a local Agent Skill and deterministic build chain. When explicitly approved, it
+can read scoped source material, write a new local Artifact and Evidence Package, and execute
+the selected local validator. It does not provide authentication, sandboxing, secrets storage,
+remote execution, publishing or a Runtime permission system.
+
+A security report is appropriate when Loop Craft unexpectedly crosses an approved filesystem or
+authority boundary, leaks excluded source material into an Artifact, accepts tampered Evidence,
+weakens a stop/approval condition, or makes an unsupported compatibility claim. General feature
+requests and ordinary classification disagreements belong in the public Issue tracker after all
+sensitive material is removed.