loganavter/improve-imgsli/help-authoring
Claude Code and agent skill by loganavter/improve-imgsli
- community
Checking that the SKILL.md file is still in the repository.
Authors and maintains the in-app Help manual (host + tab topics, tree.json merge, figure budgets, {{tr:}}/{{img:}} tokens, i18n keys, figure checks). Use when adding or updating help pages, contributing a tab's help subtree, changing help figures or their assets, or fixing help i18n keys.
Install
npx vskill@latest install loganavter/improve-imgsli/help-authoringSKILL.md
In-app Help authoring
Improve-ImgSLI ships an illustrated manual (hubs → cards → pages) rendered by the app's HelpDialog (src/plugins/help/dialog.py): toolkit parse_help_blocks → Text(mode="document"), with host-owned TOC card, scroll area and figure lightbox. Primary discovery stays Find Action; Help teaches shape and habits.
Authoritative reference: docs/dev/HELP_SYSTEM.md — read it before designing a page. This skill is the decision map.
When to touch Help
Per AGENTS.md, any user-visible change (behavior, hotkeys, settings, workflows) updates the matching help page in the same task. New tab → contribute its own subtree; new feature → new ### section or page; changed dialog → update its page.
Where things live
| Owner | Path | ||||
|---|---|---|---|---|---|
| Host scan / dialog | src/plugins/help/ | ||||
| Host topics + shared assets | src/resources/help/ (`en | ru | zh | pt_BR/, assets/`) | |
| Tab topics | src/tabs/<tab>/help.py (build_help_hub_contribution() for multi-page hubs, mount = workspace.<tab> by convention) + src/tabs/<tab>/resources/help/ | ||||
| Rendering | toolkit parse_help_blocks → Text(mode="document"); dialog, TOC, scroll area and lightbox are app chrome (src/plugins/help/dialog.py, image_lightbox.py) — read-only from the app side |
Tabs publish hub presentation via create_service("contribute_help"); the Help plugin never imports tabs.*. Structure itself is scanned from the filesystem — dirs are hubs, *.md files are pages, ids are dotted paths.
Authoring rules (summary)
- Page title
##; concept sections###with stable{#anchor}; the TOC is generated from anchored###— keep titles short. - Tables parse (
parse_help_blocks→TableBlock: first row is the header when the next line is a---separator;\|escapes a pipe) — but the corpus test still rejects| … |rows in shipped bodies (test_help_tree.py), so keep- **X** — …lists until that assertion is lifted. - Labels are literal, on-screen names in
**bold**(no{{tr:}}tokens): write the control exactly as the UI shows it. Bodies ship in en + ru + zh + pt_BR; a missing language falls back to EN. Prose canon: docs/dev/HELP_STYLE_GUIDE.md. <!-- doc-truth:allow --> - Figures as body-relative paths into the owning package's
assets/. Never add figures outside the per-topic budget table in HELP_SYSTEM.md. - Cross-link with
help://dotted.path#anchor; no trailing### Relateddumps — weave links into prose or use a short### Next topics.
Host vs tab content
- Keep host generic pages to shared behavior only; put tab specifics in contributed tab pages.
- In host leave a stub only:
###header + stable{#anchor}+ one line +help://link to the tab overview. - Keep anchors stable — TOC and
help://dotted.path#anchorlinks depend on them; never rename to fix duplication. - Anti-pattern (pre-webcanon): a host page duplicated the MC overview instead of linking to
help://workspace.multi_compare.overview. - Positive pattern:
platform.settingsshell + tab sections viabuild_settings_contribution(src/tabs/image_compare/use_cases/registration.py:49).
Figure policy (default zero)
A page does not need a picture. Add :::figure only when a screenshot answers a question prose cannot (layout, spatial gesture, dialog shape, a mode that looks different).
| Budget | When | Placement | |
|---|---|---|---|
| 0 | Chord/settings inventories, anything fully named by UI strings | — | |
| 1 | One primary surface / spatial idea for the whole page | after intro, prefer side=block | |
| 2 | Page teaches distinct visual states | one figure next to the ### it illustrates | |
| 3+ | Exception: several modes/gestures that each look different | one per mode/gesture section |
Rules: need before asset; screenshots go to final filenames under the owning package's assets/; caption explains the picture (what is shown), not a menu path; size with width=/height= (prefer height= for toolbar strips); lightbox is free.
Verification loop
Key tests: test_help_tree.py (paths, hub fragments, scenario-card page shape), test_help_bodies.py (literal labels — no {{tr:, body-relative images), <!-- doc-truth:allow --> test_help_browser_navigation.py (incl. Missing-page), test_help_dialog_opens.py. Scenario shape (## + ###) is enough for the tests — figures are never required.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/help-authoring/SKILL.md
Popularity
- Repository stars
- 45
- Repository forks
- 5
- 7-day trend score
- 0
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/loganavter/improve-imgsli/help-authoring)