Code Wiki
SkillDocs & knowledgeBuilds a local CodexQA architecture knowledge graph from community detection and `wiki inputs` (no model), then writes a DeepWiki-style HTML wiki report. Use when the user mentions codexqa-code-wiki, code-wiki, wiki, knowledge graph, architecture wiki, module map, reading guide, 代码知识图谱, 架构 Wiki, 模块地图, 阅读导览, wiki inputs, --no-llm, community detection, 社区检测, HTML 报告, or asks to map modules / generate a repo wiki without an LLM. Not change review (that is codexqa-code-analyzer), not CodexQA evidence-pack HTML review (that is codexqa-code-reviewer), and not SAST+agent code-risk scan reports (that is codexqa-defect-analyzer). Former skill name: code-wiki.
Available today. Use it from your connected AI after setup.
No other account needed.
Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.
Then ask your AI: use the Code Wiki skill
What this skill tells your AI
The instructions your AI receives, as published by openqa-cn/codexqa in skills/codexqa-code-wiki/SKILL.md and read by ahel’s review.
Local architecture knowledge graph: index first, then export community
digests with wiki inputs and write a DeepWiki-style HTML wiki
report (sidebar + article + on-this-page TOC, module map, reading guides).
There is no codexqa wiki without --no-llm in this skill, and no
wiki embed / query wiki. Those call a model or need an embedded wiki.
The graph facts are already in wiki inputs JSON:
index → wiki inputs (architecture / overview / page / visualization) → HTML report
Pick the scenario before acting. Do not run wiki commands until an index
exists. Do not send system / user prompt strings to another model.
README.md / README.zh-CN.md are human-facing. Do not load them at runtime.
Documents (load on demand)
Read this file first. Read another file only when the row below applies. Do not preload the whole tree.
| File | Load when |
|---|---|
SKILL.md (this file) | always: routing, report contract, reject conditions |
| references/playbook.md | entering a scenario (index health / map / module / guide / persist) |
| references/report.md | before writing the HTML file |
| assets/report-template.html | copy this file; do not invent a new layout |
| references/diagrams.md | before drawing; copy init and classDef verbatim |
| references/cli.md | CLI missing, PATH, or maintenance |
README.md, README.zh-CN.md | human-facing; not needed by the agent |
Scenario routing
Open references/playbook.md and jump to the named section.
| User is asking… | Playbook section |
|---|---|
| What is this repo / module map / knowledge graph | Map the repository (index health first) |
| How is the system organized / architecture wiki | Map the repository |
| What does this module do / who does it talk to | Explain one module |
| Where should I start reading / guided path | Reading guide |
| Persist a rule-only wiki for the UI | Persist rule-only wiki |
| Empty results / missing index / wiki failed | Index health / Wiki blockers |
Change review, callers, test gaps, and stack traces belong to codexqa-code-analyzer.
Report contract
Deliver a self-contained HTML knowledge-graph report in DeepWiki
wiki layout (left sidebar tree, article, on-this-page TOC) using the same
daytime HTML chrome as codexqa-testcase-generator (light default, 中文/EN
and 白天/黑夜 toggles, green accent #0f6b4c).
Copy assets/report-template.html
into the working directory, then fill slots with Edit (do not rewrite CSS).
Default the filled report to Simplified Chinese (headings, overview, guides, notes). Show community aliases as P01 (capitalize the leading p). Keep symbol names as-is. Write for a
newcomer: what the system is, where to start, which module is the hub,
which pages are standalone. Not a product brochure, and not an Archify /
architecture canvas.
Read references/report.md before writing the file.
Read references/diagrams.md before drawing. A diagram that misses the quality bar fails the report.
Report body (HTML slots) is only these blocks:
- How the system works (from
overview/architectureinput, in plain language) - Module map (community ids, human titles, real
deps) - Layers / data flow (same modules, grouped Entry → Storage; no invented edges)
- Reading path (only steps backed by a real dependency)
- Module notes (responsibility / public API / internal calls / cross-module traffic)
- Peripheral modules (empty
deps— do not force them into a layer) - Architecture diagram: at least one Mermaid block in
.diagram, and it must pass the quality bar - Fill the sidebar tree (
#nav-entry/#nav-app/#nav-domain/#nav-storage/#nav-modules/#nav-peripheral) so it looks like a DeepWiki wiki, not 8 flat links
Trust only the input object on each wiki inputs row. system / user are prompt templates, not evidence. page rows carry the digest; visualization / architecture / overview use rule titles and have no model-written body.
Reject the whole report and rewrite if any of these hold:
- Written as a generic project review (product intro, use cases, scored pros/cons)
- Evidence comes from README / a website / guesswork, not this run of
wiki inputs - Invented a module, group, or edge that is not in
communities/deps/cross_community/ visualizationcandidates - Put an isolated module (empty
deps) into a functional layer - Ran
codexqa wikiwithout--no-llm, or ranwiki embed/query wiki - Mermaid is missing the paper
init, the threeclassDeflines, or a core module that should beriskhas noclass ... risk - Architecture
subgraphtitles are package names (Renderer / Compiler / Shared) instead of Entry → Application → Domain → Storage - HTML is missing the bundled wiki chrome (sidebar + article + TOC), or was written from scratch instead of copying the template
- Report was delivered as Markdown-only / chat-only with no HTML file
- Archify / grouped-swimlane architecture canvas was generated (this skill does not ask for that)
- Chrome or body left in English when the user did not ask for English (overview / pages selected)
- Overview dumps field names (
deps,cross_community,wiki inputs) instead of responsibility / dependency / reading order
Signals
- GitHub stars
- 60
- Forks
- 2
- Last commit
- Sep 2026
ahel review
K1binfo
installs-packages (in README.md)K1binfo
installs-packages (in README.zh-CN.md)K1binfo
installs-packages (in references/cli.md)
Automated review, not a security audit. Ruleset v1+k2.
Advanced
- Item type
- skill
- Key
codexqa-code-wiki- Source
- github.com/openqa-cn/codexqa