AgentStack
SKILL verified MIT Self-run

Youqing Laozhao Skill

skill-qianqian5774-youqing-laozhao-skill-youqing-laozhao-skill · by qianqian5774

Research and review manga from a title and author. Use when a user asks “这本漫画怎么样”, “有请老赵”, requests a manga recommendation or review, or wants autonomous work identification, multilingual research, fact-checking, recommendation judgment, Chinese review writing, and delivery as a finished responsive webpage plus long image.

No reviews yet
0 installs
0 views
view→install

Install

$ agentstack add skill-qianqian5774-youqing-laozhao-skill-youqing-laozhao-skill

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Youqing Laozhao Skill? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

有请老赵.skill

Accept the manga title and author as sufficient input. Take responsibility for ordinary research gaps. Deliver a finished Chinese manga-review webpage and an automatically rendered long PNG by default.

User contract

  • Require only 漫画书名+作者.
  • Do not ask the user for ISBN, edition, interviews, reviews, reading notes, scans, or a research plan.
  • Accept optional audience, spoiler, edition, or tone preferences without making them prerequisites.
  • Keep research records and production metadata inside the run package.
  • Present the final webpage directly. Do not display “AI 生成”, “机器草稿”, “未经人工审核”, run IDs, confidence labels, or audit status on the page.
  • Stop before writing only when the work cannot be identified, an outcome-changing same-name ambiguity remains after research, or virtually no verifiable information exists.

Run the workflow

  1. Resolve the title, original title, creator roles, series, volume scope, region, language, edition, reprint, and adaptation relationships.
  2. Research the original language, Chinese, and English when useful. Prefer primary and authoritative sources; add professional criticism and community reactions without treating them as equivalent evidence.
  3. Separate facts, work evidence, interpretations, reader reactions, and uncertainties. Track provenance and counterevidence.
  4. Generate at least two recommendation claims. Select one only after testing evidence coverage, reader value, limitations, version effects, and subject attribution.
  5. Read annotations/style-genes/candidate-library-v0.2.json. Select 3–6 applicable genes across perspective, argument, and paragraph layers. Its bundled source assets preserve method provenance; load them only when a gene’s rationale needs inspection.
  6. Read references/chinese-writing-standard.md. Compose from Chinese semantic habits. Avoid translated argument scaffolding, binary contrast templates, colon-heavy explanations, and report-style transitions.
  7. Write a complete natural review. Include a clear judgment, manga-specific analysis, useful context, real limitations, reader fit, and a conclusive recommendation.
  8. Run both prose audits, build the webpage and long PNG in one delivery command, validate the run, and inspect both outputs.

Use this sequence when a shell is available:

python scripts/review_pipeline.py init --title "" --author "" --run-kind production
python scripts/audit_chinese_prose.py exports/runs//run.json
python scripts/audit_style_similarity.py exports/runs//run.json
python scripts/build_review_delivery.py exports/runs//run.json
python scripts/review_pipeline.py validate exports/runs//run.json

Before the first long-image export, ensure the runtime is ready with python -m pip install -r requirements-long-image.txt and python -m playwright install chromium. If Chromium is absent on Windows, the exporter can use Microsoft Edge. Adapt python to the environment’s Python 3.9+ executable when necessary.

Research rules

  • Open and inspect sources. Do not use search-result snippets as final evidence.
  • Use official and bibliographic sources for identity, edition, dates, publication facts, and creator roles.
  • Use accessible work material for scenes, visual form, paneling, structure, and dialogue.
  • Attribute professional interpretations and community reactions to their actual subjects.
  • Remove, narrow, or attribute claims that remain uncertain after reasonable cross-checking.
  • Do not transfer anime cinematography, motion, music, voice acting, or reception to the manga.
  • Do not fabricate full-reading experience, page numbers, quotations, scenes, or visual evidence.

Read references/workflow.md when resolving evidence, versions, research coverage, recommendation claims, or exceptional stops. Read references/runtime-contract.md when handling provenance, publishing boundaries, metadata, images, or uncertainty.

Writing rules

  • Start from a concrete person, action, scene, object, public impression, or reading problem when the evidence supports it.
  • Let judgment arrive through evidence and paragraph movement.
  • Return every external detour to the manga.
  • Make limitations change the recommendation rather than appearing as a ceremonial disclaimer.
  • Default to zero “不是……而是……” and close variants.
  • Keep Chinese-colon density at or below one per thousand Chinese characters.
  • Default to zero “首先/其次/最后”, “值得注意的是”, and conclusion scaffolding.
  • Never inherit personal experiences or distinctive phrases from calibration materials.
  • Keep normal source citations while removing production meta from the finished page.

Final delivery

Deliver both exports/runs//index.html and one exports/runs//review-long.png, not merely chat prose or Markdown. The PNG uses the dedicated long-image layout at 1080px width, with a 22px body and a 920px article container; it hides interactive controls, waits for fonts and images, captures in safe slices, and joins them without seams. If the page exceeds the safe single-file height, deliver numbered PNG parts recorded in run.json. The webpage must provide a same-tab long-image preview and a direct download link to the actual PNG or all numbered parts; rerender the final HTML after PNG export so the links stay current. Ensure readable title hierarchy, body width, type size, line height, spacing, images with provenance and alt text, references, mobile layout, and print layout. Open or show both outputs automatically when the environment supports local-page viewing.

If the environment cannot open a local file automatically, provide the absolute path and tell the user it can be opened directly in a browser. Do not turn that environment limitation into a research question for the user.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.