majiayu000/spellbook / skills/xray
xray
VerifiedInvestigate how a concept, code path, application, network flow, system, incident, document, or local artifact actually works, then deliver a two-depth visual HTML explainer with a dead-simple big-picture first layer and source-backed technical depth behind it. Use when the user invokes $xray, asks how something works under the hood, wants a module or app behavior traced across boundaries, requests web, code, network, or safe static artifact research before explanation, or asks for an evidence-backed explainer. Do not use for answer-only definitions, general web design, full codebase audits, or invasive binary reverse engineering without explicit authorization.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 13.2 KB |
| agents/openai.yaml | config | 379 B |
| assets/concept-explainer.html | other | 9.0 KB |
| assets/feature-explainer.html | other | 7.9 KB |
| evals/evals.json | config | 7.9 KB |
| references/chinese-writing.md | markdown | 3.8 KB |
| references/code-archaeology.md | markdown | 4.7 KB |
| references/evidence-contract.md | markdown | 3.2 KB |
| references/research-routing.md | markdown | 2.8 KB |
| references/reverse-core.md | markdown | 8.2 KB |
| references/visual-explanation.md | markdown | 5.1 KB |
| scripts/render-check.shexecutable | script | 1.9 KB |
references/chinese-writing.md
View on GitHub# Xray Chinese Writing Pass
Load this reference only after the evidence and causal model are stable. It is a task-specific projection of human-writing principles for Chinese technical explainers, not a general writing skill.
## Preserve the Ground Truth
- Keep exact versions, symbols, addresses, field names, error strings, quotations, dates, uncertainty, and source meaning unchanged.
- Do not turn a static clue into reachable behavior, an inference into an observation, or an unknown into a smooth narrative bridge.
- Do not invent personal experience, dialogue, scenes, user reactions, or precise details to make the explainer feel human.
- If the material cannot support a claim, shorten the explanation or keep the gap visible. Repetition does not create evidence.
## Give the Reader a Speaking Position
Write like someone who inspected the target and is now walking a curious reader through what matters.
- Say what the mechanism does early. Do not preview the report structure or research process.
- Let concrete actions carry the prose. Name who reads, decides, transforms, stores, sends, retries, or fails.
- Put evidence near the claim it supports. Explain why an address, request field, log line, or persisted record matters instead of merely listing it.
- Admit uncertainty once, at the point where it changes the explanation. Do not repeat caution labels in every paragraph.
## Keep the Two Depths Distinct
### Orientation layer
- Answer the teaching question in ordinary Chinese before introducing technical inventory.
- Use the smallest accurate visual and enough words to explain its input, central decision or transformation, output, and important failure boundary.
- Prefer familiar verbs over abstract nouns. Translate jargon when possible, but preserve a technical literal when changing it would make the explanation inaccurate.
- Do not let hashes, package formats, provider names, or methodology dominate the first screen unless one of them is the teaching question.
### Evidence layer
- Preserve exact anchors and direct source links.
- Explain one reachable path before showing peripheral clues.
- Separate observed evidence, corroborated behavior, inference, and unknowns only where the distinction matters.
- Keep paragraphs and cards unequal when the evidence is unequal. Do not force every step into the same sentence shape or word count.
## Remove Model-Like Chinese
- Delete report-like headings, repeated summaries, fake suspense, inflated conclusions, and generic insight phrases.
- Avoid commercial and institutional jargon when a person, action, cost, delay, or consequence can say the same thing plainly.
- Do not manufacture a misconception merely to overturn it. State the conclusion directly and put the evidence nearby.
- Avoid three or more parallel clauses built from the same template. Keep a list only when the reader genuinely needs a list.
- Do not use colons or em dashes as rhetorical scaffolding in headings, captions, or prose. URLs, code, machine fields, and faithful technical literals are exempt.
- Do not end by summarizing the whole page again or enlarging a bounded mechanism into a claim about an industry, era, or future.
## Read It Aloud Once
On the final pass:
1. Find sentences whose subject or action arrives too late and move the actor forward.
2. Break repeated paragraph rhythms. Let a dense causal step take more room and a simple transition end quickly.
3. Remove any sentence that only restates the previous card or diagram label.
4. Check that pronouns still point to the correct component when client, server, runtime, and provider appear together.
5. Verify that every rewrite preserved the evidence level and citation relationship.
Use judgment for this pass. Do not create a hardcoded prose validator or treat punctuation counts as proof that the writing sounds human.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |