External Agent Skills To Adapt

External Agent Skills To Adapt

This page ranks which external high-install/high-signal skills should be adapted into Hermes workflows, kept as reference material, or ignored for now. The default stance is do not install blindly.

Decision framework

Classify external skills into four buckets:

1. Adapt now — immediately useful, aligns with Hermes workflows, low platform lock-in.
2. Adapt later — good pattern, but requires more tooling or a clear user need.
3. Reference only — useful design ideas but too platform-specific or not currently relevant.
4. Ignore for now — high installs but low relevance, redundant, or mostly vendor promotion.

Criteria:

• Recurring user need.
• Clear trigger surface.
• Low dependency mismatch with Hermes tools.
• Strong verification/fallback design.
• Low risk of hidden side effects.
• Supports the principle that main agent owns orchestration; scripts stay thin.

Adapt now

1. Agent skill discovery workflow

• Source: vercel-labs/skills/find-skills
• Raw: raw/articles/agent-skills-discovery-2026-06-19/vercel-labs__skills__find-skills.md
• Why: Directly matches the current workflow: find high-signal skills, archive them, read before installing.
• Hermes adaptation: a local skill that searches Hermes Hub + skills.sh + GitHub, ranks candidates, downloads SKILL.md, writes wiki notes, and asks before install.
• Priority: high.

2. Skill creation and evaluation harness

• Source: anthropics/skills/skill-creator
• Raw: raw/articles/agent-skills-discovery-2026-06-19/anthropics__skills__skill-creator.md
• Why: Strongest pattern for improving Hermes skills over time.
• Hermes adaptation: create a repeatable evaluation loop for local skills: trigger tests, with/without skill comparisons, qualitative grading, and patch recommendations.
• Caveat: avoid heavyweight eval machinery unless a skill is important or frequently used.
• Priority: high.

3. Architecture grilling / codebase architecture review

• Source: mattpocock/skills/improve-codebase-architecture
• Raw: raw/articles/agent-skills-discovery-2026-06-19/mattpocock__skills__improve-codebase-architecture.md
• Why: Useful for software-development sessions. It turns architecture critique into a structured exploration + candidate report + user grilling loop.
• Hermes adaptation: integrate with existing software-development skills, especially planning, code review, simplify-code, and debugging.
• Priority: high-medium.

4. Browser skill forge

• Source: browser-act/skills/browser-act-skill-forge
• Raw: raw/articles/agent-skills-discovery-2026-06-19/browser-act__skills__browser-act-skill-forge.md
• Why: Strong pattern for turning one-off website exploration into reusable automation.
• Hermes adaptation: use Hermes browser tools or terminal scripts to explore APIs/DOM, then write a reusable skill or script wrapper.
• Caveat: must be adapted carefully because browser-act CLI semantics differ from Hermes browser tools.
• Priority: medium-high.

Adapt later

Supabase operational runbook

• Source: supabase/agent-skills/supabase
• Raw: raw/articles/agent-skills-discovery-2026-06-19/supabase__agent-skills__supabase.md
• Why: Excellent security and migration discipline. Useful if Supabase appears in user projects.
• Priority: medium.

Cloudflare / Wrangler workflows

• Sources: cloudflare/skills/cloudflare, cloudflare/skills/wrangler
• Raw:

- raw/articles/agent-skills-discovery-2026-06-19/cloudflare__skills__cloudflare.md

- raw/articles/agent-skills-discovery-2026-06-19/cloudflare__skills__wrangler.md

• Why: Strong operational pattern for deployment/infrastructure tasks.
• Priority: medium.

Expo / React Native workflows

• Source: expo/skills
• Raw:

- raw/articles/agent-skills-discovery-2026-06-19/expo__skills__building-native-ui.md

- raw/articles/agent-skills-discovery-2026-06-19/expo__skills__native-data-fetching.md

- raw/articles/agent-skills-discovery-2026-06-19/expo__skills__upgrading-expo.md

• Why: Useful if mobile work recurs.
• Priority: medium-low.

Firebase workflows

• Source: firebase/agent-skills
• Raw:

- raw/articles/agent-skills-discovery-2026-06-19/firebase__agent-skills__firebase-basics.md

- raw/articles/agent-skills-discovery-2026-06-19/firebase__agent-skills__firebase-auth-basics.md

- raw/articles/agent-skills-discovery-2026-06-19/firebase__agent-skills__firebase-hosting-basics.md

• Why: Clear official runbook style, but only useful when Firebase projects exist.
• Priority: medium-low.

Web quality and accessibility review

• Sources: addyosmani/web-quality-skills, vercel-labs/agent-skills/web-design-guidelines, anthropics/frontend-design
• Raw:

- raw/articles/agent-skills-discovery-2026-06-19/addyosmani__web-quality-skills__accessibility.md

- raw/articles/agent-skills-discovery-2026-06-19/addyosmani__web-quality-skills__seo.md

- raw/articles/agent-skills-discovery-2026-06-19/vercel-labs__agent-skills__web-design-guidelines.md

- raw/articles/agent-skills-discovery-2026-06-19/anthropics__skills__frontend-design.md

• Why: Broadly useful for frontend review and design critique.
• Priority: medium.

Reference only

Azure skills cluster

• Source: microsoft/azure-skills
• Representative raw:

- microsoft__azure-skills__microsoft-foundry.md

- microsoft__azure-skills__azure-prepare.md

- microsoft__azure-skills__azure-deploy.md

- microsoft__azure-skills__azure-validate.md

- microsoft__azure-skills__azure-diagnostics.md

• Why reference: excellent examples of complex operational skill design, context resolution, preflight checks, and sub-skill routing.
• Why not adapt now: user does not appear to be actively working in Azure; high install count likely reflects ecosystem bundling.

inference.sh / media generation skills

• Source: qu-skills/skills
• Representative raw:

- qu-skills__skills__ai-video-generation.md

- qu-skills__skills__ai-image-generation.md

- qu-skills__skills__ai-avatar-video.md

- qu-skills__skills__remotion-render.md

• Why reference: useful catalog style and CLI-wrapper pattern.
• Why not adapt now: depends heavily on inference.sh account/CLI and overlaps with Hermes media tools.

Baoyu publishing/media skills

• Source: jimliu/baoyu-skills
• Why reference: potentially useful for Chinese content workflows, WeChat posting, markdown-to-HTML, image generation.
• Why not adapt now: needs explicit user workflow before installing or porting.

Ignore for now

• Platform-specific skills for tools not currently used.
• Skills whose only value is wrapping a third-party CLI without extra verification or reasoning.
• Skills that require installation of large local inference stacks, unless the user explicitly asks. The user prefers cloud APIs over local LLM runtimes.

Proposed next concrete actions

1. Create a Hermes local skill: agent-skill-discovery.

- Trigger: user asks to find/load/review external skills.

- Workflow: discover → rank → archive raw → summarize → ask before install.

1. Create or patch a Hermes skill-authoring checklist.

- Incorporate External-Agent-Skills-Design-Patterns.

- Add test prompt / trigger evaluation expectations.

1. Review existing local Hermes skills for weak descriptions.

- Prioritize agent orchestration skills and Hermes troubleshooting skills.

- Patch only where clear value exists.

1. Build a weekly skill radar cron if desired.

- Compare skills.sh leaderboard changes.

- Download newly high-growth candidates into wiki.

- Deliver a short digest rather than installing anything.

Related

• External-Agent-Skills-Design-Patterns
• Agent-Skill-Discovery
• Agent-Skills-Hub-and-Skills-sh-High-Star-Skills