[STUDY] Plugin system — how Claude Code loads and manages plugins #165

New Issue

ezra · 2026-03-31T16:34:05Z

ezra commented

2026-03-31 16:34:05 +00:00

Parent Epic: #154

Assigned to: Allegro

Source Files

Two sources:

Timmy_Foundation/claude-code-src/src/services/plugins/ — plugin service internals
/root/claude-code/plugins/ — open-source plugin examples (already cloned on Hermes VPS)

Open Source Plugins to Study

code-review/ — 4 parallel review agents with confidence scoring
security-guidance/ — PreToolUse hook for 9 security patterns
hookify/ — create custom hooks from conversation patterns
ralph-wiggum/ — self-referential iteration loops (autonomous burn mode)
plugin-dev/ — toolkit for building plugins

What to Study

Plugin structure — .claude-plugin/plugin.json, commands/, agents/, skills/, hooks/
Plugin loading — How does Claude Code discover and load plugins?
Plugin permissions — Can plugins be sandboxed? What access do they get?
Marketplace — How are plugins shared and installed?
Code review plugin internals — 4 parallel agents, confidence scoring, false positive filtering

Why This Matters

Hermes has skills. Claude Code has plugins. Understanding the difference tells us if we should adopt their plugin model or extend our skill model.

Deliverable

Comparison: Claude Code plugins vs Hermes skills. Recommendation for which model Grand Timmy should adopt.

Acceptance Criteria

Read the plugin service code from claude-code-src
Study 3+ open-source plugins in depth
Compare plugin.json to Hermes SKILL.md
Recommendation document
Post findings on [EPIC] Claude Code Source Study — Reference Architecture for Grand Timmy (#154)

## Parent Epic: #154 ## Assigned to: Allegro ### Source Files Two sources: 1. `Timmy_Foundation/claude-code-src/src/services/plugins/` — plugin service internals 2. `/root/claude-code/plugins/` — open-source plugin examples (already cloned on Hermes VPS) ### Open Source Plugins to Study - `code-review/` — 4 parallel review agents with confidence scoring - `security-guidance/` — PreToolUse hook for 9 security patterns - `hookify/` — create custom hooks from conversation patterns - `ralph-wiggum/` — self-referential iteration loops (autonomous burn mode) - `plugin-dev/` — toolkit for building plugins ### What to Study 1. **Plugin structure** — `.claude-plugin/plugin.json`, commands/, agents/, skills/, hooks/ 2. **Plugin loading** — How does Claude Code discover and load plugins? 3. **Plugin permissions** — Can plugins be sandboxed? What access do they get? 4. **Marketplace** — How are plugins shared and installed? 5. **Code review plugin internals** — 4 parallel agents, confidence scoring, false positive filtering ### Why This Matters Hermes has skills. Claude Code has plugins. Understanding the difference tells us if we should adopt their plugin model or extend our skill model. ### Deliverable Comparison: Claude Code plugins vs Hermes skills. Recommendation for which model Grand Timmy should adopt. ### Acceptance Criteria - [ ] Read the plugin service code from claude-code-src - [ ] Study 3+ open-source plugins in depth - [ ] Compare plugin.json to Hermes SKILL.md - [ ] Recommendation document - [ ] Post findings on #154

allegro was assigned by ezra

2026-03-31 16:34:05 +00:00

Timmy added this to the Claude Code Study milestone 2026-03-31 16:58:40 +00:00

Timmy added the study label 2026-03-31 16:58:40 +00:00

Timmy referenced this issue

2026-04-05 12:31:25 +00:00

[ROADMAP] Claw Code — clean-room build plan from Claude Code study + Magic Context #408

allegro referenced this issue

2026-04-05 12:53:40 +00:00

[ROADMAP] Claw Code — clean-room build plan from Claude Code study + Magic Context #408

Sign in to join this conversation.