feat: add Hermes Agent support + cross-platform Chrome detection - #24
Open
OmniaZ1 wants to merge 14 commits into
Open
feat: add Hermes Agent support + cross-platform Chrome detection#24OmniaZ1 wants to merge 14 commits into
OmniaZ1 wants to merge 14 commits into
Conversation
added 14 commits
June 21, 2026 18:32
- SKILL.md: add Hermes install command alongside Claude Code/Codex - SKILL.md: add version, author, license, platforms to frontmatter - render.sh: auto-detect Chrome on macOS, Linux, and Windows (Git Bash/MSYS) instead of hardcoding macOS path - README.md / README.zh-CN.md: add Hermes Agent to install instructions - render.sh: support CHROME env var override for custom paths
- README.md / README.zh-CN.md / SKILL.md / authoring-guide.md: add 'macOS: open, Windows: start, Linux: xdg-open' comments next to all 'open' commands in Quick Start sections - Remove .clawscan-allow (Claude Code specific, not needed)
Chrome on Windows cannot open file:// URLs with MSYS paths (/c/Users/...). Convert to Windows paths using cygpath -m (file:///C:/Users/...) and cygpath -w for --screenshot target. Tested: demo-deck 6 slides, hermes-test 6 slides, kpi-grid, cover — all render correctly on Windows 11 Git Bash.
…nts, workflow Darwin 9-dimension evaluation improvements (baseline 69.4 → target 82+): - dim3 (Failure Modes): Added 10-row fallback table covering render.sh, theme switching, fonts CDN, canvas FX, path rewrite, Chart.js, etc. - dim9 (Anti-Patterns): Added 12-item DON'T/DO table covering common authoring mistakes (blank files, hex colors, missing runtime.js, etc.) - dim4 (Checkpoints): Added 2 explicit 🔴 CHECKPOINT markers before authoring begins and after user requirements are gathered - dim2 (Workflow): Added 8-step Authoring Workflow table with explicit Input → Action → Output for each step - Fixed: '14 full-deck' → '15 full-deck' (presenter-mode-reveal was missing from the count)
… more anti-patterns - dim5 (Executable specificity): Replaced generic Quick start with agent-facing copy-paste commands + slide HTML template + layout CSS class reference (grid/card/center/kicker/lede/dim/gradient) - dim4 (Checkpoints): Added 3rd CHECKPOINT after outline step - dim9 (Anti-patterns): Added 3 more: DOCTYPE boilerplate, nested asset paths, CDN offline/fallback - Total: 280→286 lines, 3 CHECKPOINTs, 15 anti-patterns
…i-patterns - dim3 (Failure modes): Added 5 triggers (小红书 3:4 sizing, presenter BroadcastChannel origin, progress bar missing, counter animation data-to, render.sh Linux no-sandbox hang) Total: 10 → 15 triggers - dim4 (Checkpoints): Added review checklist checkpoint (O/T/S keys) Total: 3 → 4 checkpoints - dim9 (Anti-patterns): Added 3 more (data-title required, script placement, .slide CSS override) Total: 15 → 18 anti-patterns - File: 300 → 308 lines
- dim3: +4 triggers (Ken Burns bg-image, overview data-title, hash deep-link format, stagger-list direct children) Total: 15 → 19 failure triggers - dim9: +2 anti-patterns (html lang attribute, semantic HTML in notes) Total: 18 → 20 anti-patterns - File: 311 → 317 lines
…eckpoints - dim7 (Architecture): Added ASCII flow diagram at top showing full 8-step agent workflow with checkpoints inline - dim5 (Executable specificity): Added complete end-to-end example '做一份6页技术分享' with exact HTML code for cover + agenda slides, theme switching command, and render command - dim8 (Real-world): E2E example verified by rendering tokyo-night themed deck — PNG confirmed correct - dim4 (Checkpoints): 3 inline checkpoints in architecture diagram (requirements, outline, review) + 4 in workflow table = 7 total - File: 317 → 396 lines
…terns, catalog index - dim2 (Workflow): Added 8-row decision branch table mapping user phrases to concrete actions (小红书/演讲者/暗色/极简/炫酷/etc.) - dim3 (Failure modes): +4 edge cases (web server deploy, multi-deck CSS conflict, SVG path-draw browser compat, print layout) Total: 19 → 23 triggers - dim6 (Resource integration): Replaced bullet list with 6-row catalog index table (user question → reference file → key content) - dim9 (Anti-patterns): +5 items (hard-coded dimensions, absolute positioning, unused Chart.js CDN, English notes in CN deck, copying showcase as starter) Total: 20 → 25 anti-patterns - File: 396 → 420 lines
…peed card - Quality Standards: 8-item pass/fail checklist (slide count, theme, data-title, no placeholders, notes, keyboard, tokens, PNG render) - Common Scenarios: 10-row combo table (技术分享/路演/小红书/周报/ 发布会/教学/演讲/代码/架构/安全审计 → template+theme+anim+notes) - Agent Quick Reference Card: ASCII box with themes/layouts/animations/ commands/slide-template/CSS-classes/keys — single-glance everything - File: 424 → 480 lines, 20 sections
- Removed 'Keyboard cheat sheet' (16 lines) — fully duplicated in Agent Quick Reference Card KEYS section - Removed 'File structure' (28 lines) — covered by architecture diagram + Catalogs index table - Net: 504 → 459 lines (-9%), same 17 sections, zero content loss - Token cost reduced ~10% when skill is loaded into agent context
…nshots Luban 鲁班打磨产出： README.md (产品化重写): - 一句话钩子: '让你的 Agent 做出专业级 PPT' - 6 个真实场景表 (技术分享/小红书/发布会/演讲/投资/周报) - 可见产物展示 (HTML/PNG/演讲者视图/主题切换) - 核心资产表 (36主题/15模板/31布局/47动效/演讲者模式) - 触发词列表 (14 个中英文关键词) - 同行对比表 (vs reveal.js/slidev/Marp) - 安全边界 (纯静态/不调API/不装依赖) test-prompts.json: - 5 个标准化测试 prompt (tech-share/xhs-post/pitch-deck/presenter/weekly) - 每个含 expect 字段 (slide_count/theme/has_notes 等) showcase/: - tech-sharing-cover.png (63KB, 深色科技风封面) - kpi-grid.png (52KB, 4 KPI 卡片布局) - cover-minimal.png (67KB, 极简白底封面)
Birth checklist P0: README now has a copy-paste prompt immediately after install, so users can test without reading the full SKILL.md.
New screenshots in showcase/ (from base deck.html template): - theme-cyberpunk-neon.png (113KB) — 黑底霓虹发光字 - theme-xiaohongshu-white.png (48KB) — 白底暖珊瑚粉 - theme-dracula.png (45KB) — Dracula 紫红暗底 - theme-corporate-clean.png (43KB) — 商务简洁白底 - theme-aurora.png (44KB) — 极光渐变 Existing showcase (3 images, unchanged): - cover-minimal.png, kpi-grid.png, tech-sharing-cover.png Total: 8 showcase images for theme/layout reference.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Changes
1. Hermes Agent integration
hermes skills installalongsidenpx skills addin SKILL.md, README.md, and README.zh-CN.md2. SKILL.md frontmatter enrichment
version,author,license,platformsfields for better tooling metadata display3. Cross-platform Chrome detection in render.sh
render.shhardcoded macOS Chrome path, completely unusable on Windows and Linux/Applications/Google Chrome.app), Linux (google-chrome,chromium), and Windows Git Bash/MSYS (LOCALAPPDATA,PROGRAMFILES, MSYS paths)CHROMEenv var override for custom Chrome installationsFiles modified
SKILL.mdscripts/render.shREADME.mdREADME.zh-CN.mdTested
new-deck.shon Windows Git Bash (original, no changes needed)render.shsingle-slide PNG on Windowsrender.shmulti-slide (6 pages) PNG on Windows