Skip to content

chore: complete t062 research on vercel-labs/agent-skills #559

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
marcusquinn merged 2 commits into from
Feb 8, 2026
Merged

chore: complete t062 research on vercel-labs/agent-skills #559

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
999c58f
chore: mark t071 blocked in TODO.md
marcusquinn Feb 8, 2026
c169c32
chore: complete t062 research on vercel-labs/agent-skills for aidevop…
marcusquinn Feb 8, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 5 additions & 4 deletions TODO.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -264,7 +264,7 @@ Tasks with no open blockers - ready to work on. Use `/ready` to refresh this lis
- [x] t070 Backlink & Expired Domain Checker subagent #seo #domains ~15m (ai:10m test:3m read:2m) logged:2026-01-24 completed:2026-02-07
- Notes: seo/backlink-checker.md created (106 lines). Covers Ahrefs/DataForSEO backlink APIs, WHOIS expired domain detection, reclamation workflow, and integration with existing SEO subagents.
- [ ] t071 Voice AI models for speech generation and transcription #tools #voice #ai ~30m (ai:20m test:5m read:5m) ref:GH#502 logged:2026-01-24 related:t027
- Notes: Add support for voice AI models covering both TTS (speech generation) and STT (transcription). API providers: Hugging Face Inference API (TTS/STT endpoints), ElevenLabs, OpenAI TTS/Whisper. Local models: Qwen3-TTS (0.6B/1.7B, Apache-2.0, 10 languages, voice clone/design/custom, streaming, vLLM support - https://github.com/QwenLM/Qwen3-TTS), Whisper (transcription), Bark, Coqui TTS. Create subagent at tools/voice/ or tools/ai/voice.md covering: model selection (local vs API, quality vs speed), installation (pip install qwen-tts, HF download), usage patterns (TTS generation, voice cloning, voice design, transcription), streaming support, GPU requirements. Related to t027 (hyprwhspr speech-to-text).
- Notes: Add support for voice AI models covering both TTS (speech generation) and STT (transcription). API providers: Hugging Face Inference API (TTS/STT endpoints), ElevenLabs, OpenAI TTS/Whisper. Local models: Qwen3-TTS (0.6B/1.7B, Apache-2.0, 10 languages, voice clone/design/custom, streaming, vLLM support - https://github.com/QwenLM/Qwen3-TTS), Whisper (transcription), Bark, Coqui TTS. Create subagent at tools/voice/ or tools/ai/voice.md covering: model selection (local vs API, quality vs speed), installation (pip install qwen-tts, HF download), usage patterns (TTS generation, voice cloning, voice design, transcription), streaming support, GPU requirements. Related to t027 (hyprwhspr speech-to-text). BLOCKED: Max retries exceeded: clean_exit_no_signal

@coderabbitai coderabbitai Bot Feb 8, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟡 Minor

Clarify the BLOCKED reason with actionable context.

The blocker "Max retries exceeded: clean_exit_no_signal" is cryptic. For future troubleshooting, specify:

  • What operation is being retried (model download? API call? subprocess execution?)
  • Which component failed (whisper.cpp? qwen-tts? API endpoint?)
  • What clean_exit_no_signal indicates

Example: BLOCKED: Qwen3-TTS installation retry loop fails with clean_exit_no_signal after 3 attempts (suspect macOS arm64 compatibility)

🤖 Prompt for AI Agents 
In `@TODO.md` at line 267, Replace the cryptic blocker line "BLOCKED: Max retries
exceeded: clean_exit_no_signal" in the TODO entry with a precise, actionable
description that states (1) which operation was retried (e.g., model download,
API request, subprocess run), (2) which component or tool failed (e.g.,
Qwen3-TTS installer, Whisper transcription subprocess, HF inference API), (3)
how many retries occurred and any relevant platform/env (e.g., macOS arm64,
GPU/no-GPU), and (4) what "clean_exit_no_signal" means in this context (e.g.,
subprocess exited with code 0 but produced no output, or process terminated
without signal). For example, update the BLOCKED line for the Qwen3-TTS task
(tools/voice/ or tools/ai/voice.md references) to: "BLOCKED: Qwen3-TTS
installation retry loop failed after 3 attempts on macOS arm64 - installer
subprocess exited cleanly with no output (clean_exit_no_signal) while
downloading wheels" so future debuggers know operation, component, retries,
platform, and interpretation of the error.

- [ ] t072 Audio/Video Transcription subagent #tools #voice #transcription #ai ~1.5h (ai:45m test:30m read:15m) ref:GH#503 logged:2026-01-24 related:t071,t027
- Notes: Create transcription subagent at tools/voice/transcription.md supporting multiple input sources and transcription backends. **Input sources:** YouTube (via yt-dlp audio extraction), direct media URLs (audio/video files on websites), local audio files (wav/mp3/flac/ogg/m4a), local video files (mp4/mkv/webm - extract audio track via ffmpeg). **Local models (Whisper family via whisper.cpp or faster-whisper):** Tiny (75MB, speed:9.5, accuracy:6.0-6.5), Base (142MB, speed:8.5, accuracy:7.2-7.5), Large v2 (2.9GB, speed:3.0, accuracy:9.6), Large v3 (2.9GB, speed:3.0, accuracy:9.8), Large v3 Turbo (1.5GB, speed:7.5, accuracy:9.7 - RECOMMENDED default), Large v3 Turbo Quantized (547MB, speed:7.5, accuracy:9.5). **Local models (other):** NVIDIA Parakeet V2 (474MB, English-only, speed:9.9, accuracy:9.4), NVIDIA Parakeet V3 (494MB, multilingual, speed:9.9, accuracy:9.4, experimental), Apple Speech (macOS 26+, on-device, multilingual, built-in). **Cloud APIs:** Groq Whisper Large v3 Turbo (accuracy:9.6, lightning-speed inference), ElevenLabs Scribe v1 (accuracy:9.8), ElevenLabs Scribe v2 (accuracy:9.9 - highest), Deepgram Nova (accuracy:9.5, cost-effective), Deepgram Nova-3 Medical (accuracy:9.6, English-only, clinical), Mistral Voxtral Mini (accuracy:9.7, SOTA), Gemini 2.5 Pro (accuracy:9.6, multimodal), Gemini 2.5 Flash (accuracy:9.4, low-latency), Gemini 3 Pro (accuracy:9.7), Gemini 3 Flash (accuracy:9.5), Soniox stt-async-v3 (accuracy:9.6). **Helper script:** transcription-helper.sh with commands: transcribe (source → text), models (list/download local models), configure (set API keys for cloud providers). **Workflow:** detect source type → extract audio if video → select model (local/cloud based on config) → transcribe → output text/SRT/VTT. **Dependencies:** yt-dlp (YouTube), ffmpeg (audio extraction), faster-whisper or whisper.cpp (local inference). Ref: tryvoiceink.com model ratings.
- [ ] t073 Document Extraction Subagent & Workflow #plan → [todo/PLANS.md#2026-01-25-document-extraction-subagent--workflow] ~1h (ai:30m test:20m read:10m) ref:GH#504 logged:2026-01-25
Expand Down Expand Up @@ -370,8 +370,9 @@ Tasks with no open blockers - ready to work on. Use `/ready` to refresh this lis
- Notes: Git-compatible VCS (24.7k stars) with unique features: working-copy-as-commit (no staging area), operation log with undo, conflicts as first-class objects, automatic rebase of descendants, anonymous branches. Could simplify worktree workflows, enable safer experimentation, and provide better conflict handling. Evaluate: 1) Integration with existing git workflows 2) Colocated mode compatibility 3) Benefits for AI-assisted development 4) Learning curve vs productivity gains.
- [x] t061 Create debug-opengraph and debug-favicon subagents #tools #seo #browser ~10m (ai:8m test:1m read:1m) logged:2026-01-14 ref:https://opengraphdebug.com/ completed:2026-02-07
- Notes: Already implemented. seo/debug-opengraph.md and seo/debug-favicon.md both exist in .agents/seo/.
- [ ] t062 Research vercel-labs/agent-skills for inclusion in aidevops #research #tools #deployment ~10m (ai:8m read:2m) ref:GH#515 logged:2026-01-14 ref:https://github.com/vercel-labs/agent-skills
- Notes: Vercel's Agent Skills collection (332 stars, MIT). Currently includes vercel-deploy skill for instant deployment without auth. Auto-detects 40+ frameworks, returns preview URL + claim URL. Follows agentskills.io format (SKILL.md structure). Evaluate: 1) Add vercel-deploy as aidevops skill 2) Adopt Agent Skills format for aidevops skills 3) Contribute aidevops skills back to ecosystem.
- [x] t062 Research vercel-labs/agent-skills for inclusion in aidevops #research #tools #deployment ~10m (ai:8m read:2m) ref:GH#515 logged:2026-01-14 ref:https://github.com/vercel-labs/agent-skills completed:2026-02-08
- Notes: Vercel's Agent Skills collection (19.4k stars, MIT, 117 commits, 17 contributors). Now 5 skills: react-best-practices (57 rules, 8 categories), web-design-guidelines (100+ rules), react-native-guidelines (16 rules), composition-patterns, vercel-deploy-claimable. Follows agentskills.io format (SKILL.md structure). agentskills.io is now an open standard adopted by 27+ tools (Claude Code, Cursor, Gemini CLI, OpenCode, VS Code, GitHub Copilot, etc.).
- Decision: 1) vercel-deploy-claimable: SKIP - designed for claude.ai/Desktop conversations, not CLI agents. aidevops already has full Vercel CLI integration (tools/deployment/vercel.md + vercel-cli-helper.sh) which is more capable. 2) react-best-practices: CANDIDATE for import via `/add-skill vercel-labs/agent-skills` - 57 rules would complement existing code-standards. Create follow-up task. 3) Agent Skills format: ALREADY ADOPTED - aidevops already generates SKILL.md stubs via generate-skills.sh and imports external skills via add-skill-helper.sh. Full compatibility exists. 4) Contribute back: DEFERRED - aidevops skills use progressive disclosure pattern with subagent folders, which is compatible but more complex than typical SKILL.md. Consider contributing simpler skills (mermaid-diagrams, seo-audit) when the ecosystem matures.
- [x] t064 Add seo-regex subagent with Search Console regex workflow #seo #tools ~5m (ai:4m read:1m) logged:2026-01-15 completed:2026-02-07
- Notes: Created seo/seo-regex.md. Covers GSC RE2 regex patterns for brand/non-brand, questions, intent classification, long-tail, page filters, URL analysis, keyword grouping.
- [x] t083 Create Bing Webmaster Tools subagent #seo #tools ~15m actual:10m (ai:10m test:3m read:2m) logged:2026-01-29 completed:2026-02-06 related:seo-audit-skill
Expand Down Expand Up @@ -624,7 +625,7 @@ t057,Memory consolidation and pruning,,memory|optimization,2h,1h,45m,15m,2025-01
t058,Memory Auto-Capture,,plan|memory|automation|context,1d,6h,4h,2h,2026-01-11T12:00Z,pending,,,
t060,Research jj (Jujutsu) VCS for aidevops advantages,,research|git|tools,15m,10m,,5m,2026-01-13T00:00Z,pending,,,
t061,Create debug-opengraph and debug-favicon subagents,,tools|seo|browser,10m,8m,1m,1m,2026-01-14T00:00Z,done,,,
t062,Research vercel-labs/agent-skills for inclusion in aidevops,,research|tools|deployment,10m,8m,,2m,2026-01-14T00:00Z,pending,,,
t062,Research vercel-labs/agent-skills for inclusion in aidevops,,research|tools|deployment,10m,8m,,2m,2026-01-14T00:00Z,done,,,
t064,Add seo-regex subagent with Search Console regex workflow,,seo|tools,5m,4m,,1m,2026-01-15T00:00Z,pending,,,
t067,Optimise OpenCode MCP loading with on-demand activation,,opencode|performance|mcp,4h,2h,1h,1h,2026-01-21T00:00Z,done,t056,,
t069,Fix toon-helper.sh validate command,,bugfix,15m,10m,2026-01-24T00:00Z,,2026-01-25,done
Expand Down
Loading