manga-review
Skill169 starsChanged 17 days agoMITView on GitHub
Independently review manga production artifacts without editing: source text, script, storyboards, panel artwork, lettering, and final pages or PDF. Use when asked to review manga work, verify a production stage, or run an independent quality check. Keywords: review, inspect, verify, 検品, レビュー, 確認.
In this repository4 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- manga-art-directionSkill
- manga-productionSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Do not edit the production.
- Distinctive: line 14
Never claim visual inspection from a command result, hash, or file listing.
- Distinctive: line 15
If the source text was revised while you review, keep the observed version and the latest version separate.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Manga Review
You are a reviewer, not the producer. Read AGENTS.md, REVISION-AND-REVIEWS.md, and PROMPTS.md before starting.
Rules
- Do not edit the production. Review only the version and files listed in the prepared request you were given.
- Never claim visual inspection from a command result, hash, or file listing. Report what you actually read and saw, including unreadable files.
- If the source text was revised while you review, keep the observed version and the latest version separate. Never attach new page numbers or hashes to old observations.
- Include your actual runtime task identity in the response. If it is unavailable, say so; do not invent one.
Flow
The producer prepares requests and records your answers with the stage-review CLI. You only read and return JSON.
- Receive a prepared request (frozen input files plus scope) and read exactly what it lists.
- For storyboard, lettering, and final under the two-phase protocol: first read only the unnumbered images as an ordinary reader and return the natural reading order as the first-reading answer. The producer saves it with
record-reading, then sends a comparison request with the source, script, and intended order for your second answer. Without the two-phase protocol, answer the single comparison request directly. - For script review: first read only what remains in the manga (dialogue and action in order) and check whether each line's motive and response hold without the source. Then compare against the source text.
- For artwork review: inspect each panel's actual PNG against its references for identity, acting, hands, props, and continuity with the scene. Judge only visible content; never fill gaps from the script.
- Return observations, issues, and your task name as JSON. If issues are unresolved, return
changes-requested. After fixes, the producer sends a new request; never reuse an old pass as approval of new content.
Response shape
- Script pass:
dialogueReviewwith one concrete source-comparison observation per panel, including silent panels (action and conversation links). - Storyboard, lettering, and final passes:
readingReviewwith the unnumbered first reading, predicted order, and transition observations. - Artwork pass: one observation per panel covering identity, acting, hands, props, and continuity against the references.
- OCR output is a text-matching aid only. It never replaces judgment of dialogue selection, reading order, or art.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

