refactor: apply claude-code architectural patterns to simplify SuperClaude - #120
Merged
Conversation
Inspired by analyzing ~/Desktop/claude-code's architecture: Pattern 1 (Filesystem IS the Registry): - Delete agents/DEPRECATED/ (240+ files) — superseded by tiered system - Delete agents/index.yaml — self-declared deprecated - Delete config/agents.yaml — redundant with agents/core/*.md files - Remove agent list from config/superclaud.yaml — filesystem handles discovery - Remove agent trigger duplication from superclaud.yaml — defined in frontmatter Pattern 2 (One Schema, Not Six Configs): - Delete config/mcp.yaml — self-declared "documentation only" - Strip dead keys from .claude/settings.json (superclaude, skills, guardrails blocks) - Remove stale config/models.yaml reference from superclaud.yaml Pattern 3 (Guardrails at Invocation): - enforced_guardrails in settings.json was never read by any code - Real enforcement already works via PreToolUse hooks (the correct pattern) Pattern 6 (Delete Dead Code): - Remove dashboard, snake-game, hangman-game crates from Cargo.toml workspace - Delete their source directories All 1981 tests pass. No functional changes. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pattern 5 (project-specific docs only): remove content that costs context window on every conversation without providing project-specific value. AGENTS.md: Replace generic "Repository Guidelines" (coding style, testing guidelines, commit conventions) with actual agent system documentation (3-tier architecture, frontmatter schema, agent creation guide). Generic conventions are handled by ruff/linters, not LLM context. CLAUDE.md: Remove duplicate GitNexus block (100 lines) already included via @AGENTS.md. Now 35 lines of unique PAL MCP content only. Delete template rule files (.claude/rules/architecture-reference.md, logging.md, project-conventions.md) — placeholder content consuming context on every session with no project-specific value. Combined context reduction: ~340 lines → ~211 lines (-38%). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pattern 6 (delete dead code, don't archive it): remove modules that are only imported by their own tests and superseded by core/ implementations. SuperClaude/Orchestrator/ (7 files, ~2,078 LOC) — async SDK-based loop runner replaced by core/loop_orchestrator.py (sync, signal-based). Active entry point is .claude/skills/sc-implement/scripts/loop_entry.py which imports from core/, not Orchestrator/. SuperClaude/Telemetry/ (3 files, ~300 LOC) — JSONL telemetry client only imported by dead Orchestrator/events_hooks.py. Active metrics system is core/metrics.py with pluggable callback interface. Also removes tests/orchestrator/ and tests/services/test_telemetry.py (tested the dead modules). Remaining tests: 1730 passed, 0 failures. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pattern 2 (one schema, not six config files): remove config files that no code reads and simplify the one that Rust crates actually consume. Deleted: - config/quality.yaml — self-declares "semantic references for LLM", not read by any Python or Rust code. Quality weights already in superclaud.yaml (which Rust crates DO read). - config/models.yaml — self-declares "for reference". PAL MCP's listmodels provides model info at runtime. - config/consensus_policies.yaml — only referenced in README. Simplified config/superclaud.yaml from 241 to 62 lines: - Removed: commands block, dynamic_loading, logging, performance, security, development, token_optimization, triggers, workflows, environment sections — none parsed by Rust SuperClaudeConfig struct. - Kept: version, name, modes, agents, quality, worktree, mcp_servers (the fields actually consumed by crates/superclaude-core). Updated README.md to remove references to deleted files. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pattern 4 (transparent wrappers) + Pattern 6 (delete dead code): remove the SuperClaude/Agents/ package and its broken CLI consumer. The CLI `agent` command imported non-existent `AgentLoader`, so AGENT_AVAILABLE was always False — the command never worked. Similarly, scripts/report_agent_usage.py imported non-existent `usage_tracker`. Canonical agent selection lives in .claude/skills/sc-implement/scripts/select_agent.py (standalone, 551 LOC, JSON I/O) — used by the actual skill system. Deleted: - SuperClaude/Agents/ (registry.py, selector.py — 672 LOC) - setup/cli/commands/agent.py (broken CLI command) - scripts/report_agent_usage.py (broken imports) - tests/agents/ (tested dead code) - tests/scripts/test_report_agent_usage.py Updated SuperClaude/__main__.py to remove "agent" from CLI dispatch. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
|
Important Review skippedToo many files! This PR contains 299 files, which is 149 over the limit of 150. ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (299)
You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
|
✅ README Quality Check: 83/100 Structure Consistency: 100/100 See the Actions tab for the detailed report. |
Contributor
GitNexus Impact Analysis⚪ NONE Overall Risk Level
Per-File Impact
Affected ProcessesAffected ModulesDetailed Impact by FileGenerated by GitNexus impact analysis |
Auto-generated update: 8994→6382 symbols (-29%) reflecting the removal of dead modules and deprecated agents. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributor
|
✅ README Quality Check: 83/100 Structure Consistency: 100/100 See the Actions tab for the detailed report. |
Contributor
Claude Code Review (via AWS Bedrock)OverviewThis PR represents a massive architectural cleanup applying 6 patterns from Anthropic's claude-code codebase to remove 58,360 lines of dead code, redundant configuration, and broken modules. The refactor deleted 365 files while adding only 91 lines, reducing the codebase by ~29% while preserving all genuine functionality. Critical IssuesNone identified. This is exemplary technical debt cleanup. High Priority Observations✅ Pattern Application ExcellenceThe PR systematically applies 6 architectural patterns with clear rationale:
✅ Test Coverage Validation
✅ Impact Metrics
Medium Priority ObservationsDocumentation Quality
Preserved FunctionalityAll genuine innovations remain untouched:
Code Quality
Positive Observations🎯 Architectural Maturity: Learning from reference implementations (claude-code) shows engineering discipline 🎯 Risk Management: PR author ran full test suite, linting, GitNexus reindex before submitting 🎯 Transparency: Commit history shows iterative cleanup (5 focused commits vs 1 massive dump) 🎯 No Scope Creep: Resisted temptation to add features during cleanup 🎯 Evidence-Based: Identified patterns by analyzing a real codebase, not theoretical principles Review Summary
Recommendation✅ APPROVE - Merge immediately This is a masterclass in technical debt cleanup:
The only risk is not merging - every day this sits open, new code might reference the dead modules. Generated by Claude Code Review (AWS Bedrock Sonnet 4.5) - Analyzed 365 file changes, 5 commits, 6 architectural patterns |
The security-engineer agent only existed in agents/DEPRECATED/ which was deleted. The eval referenced a non-existent file, causing the Claude CLI to receive an empty prompt (bash quoting broke on the f-string double quotes). Renamed to guardian (the active core security agent) and replaced the f-string with string concatenation to avoid shell quoting issues. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributor
|
✅ README Quality Check: 83/100 Structure Consistency: 100/100 See the Actions tab for the detailed report. |
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.
Summary
Applies 6 architectural patterns learned from analyzing Anthropic's
claude-codeCLI codebase to remove dead code, redundant config, and broken modules from SuperClaude.agents/index.yaml,config/agents.yaml,config/superclaud.yamlagent lists) — filesystem discovery is the source of truthquality.yaml,models.yaml,mcp.yaml,consensus_policies.yaml), simplifiedsuperclaud.yamlfrom 241→62 linesenforced_guardrailsandguardrailsblocks fromsettings.json— real enforcement is via PreToolUse hooksSuperClaude/Agents/package (672 LOC) — broken CLI import (AgentLoadernever existed), canonical selection isselect_agent.pySuperClaude/Orchestrator/(2,078 LOC),SuperClaude/Telemetry/(300 LOC),agents/DEPRECATED/(240+ files), 3 unrelated Rust cratesImpact
What's preserved
All of SuperClaude's genuine innovations are untouched:
core/)Test plan
PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 pytest -m "not slow" tests/— 1,685 passed, 0 failuresruff check . && ruff format --check .— all clean🤖 Generated with Claude Code