Skip to content

CHANGELOG.md

Latest commit

 

History

History
442 lines (305 loc) · 27.4 KB

CHANGELOG.md

File metadata and controls

442 lines (305 loc) · 27.4 KB

Changelog

All notable changes to Claude Code Guide are documented in this file.

The format is based on Keep a Changelog.

[5.0.3] - 2026-04-24

Patch release: CC version history catches up on releases 2.1.113 through 2.1.118.

Added

  • Chapter 01 (CC Version History): Four new release entries — 2.1.118, 2.1.117, 2.1.116, 2.1.113 — covering hooks invoking MCP tools directly (type: "mcp_tool"), native binary spawn, bfs+ugrep replacing Glob/Grep, Opus 4.7 /context 1M-window fix, cleanupPeriodDays extended sweep, sandbox.network.deniedDomains, /cost+/stats→/usage merger, DISABLE_UPDATES env var, custom named themes, vim visual mode, Bash security hardening, and /resume performance gains.
  • Superseded features table: /cost and /stats commands (2.1.118) → merged into /usage (both remain as typing shortcuts).

Changed

  • Chapter 01 intro: "current 2.1.111 line" → "current 2.1.118 line".
  • "(current)" label moved from 2.1.111 to 2.1.118 in the Latest section.
  • CITATION.cff: CC 2.1.111+ compatible → CC 2.1.118+ compatible.

[5.0.2] - 2026-04-22

Patch release: Problem/Solution sub-section added at the top of Section 12 (TL;DR) across plan-mode docs, template rules, and the ExitPlanMode hook.

Changed

  • Chapter 01 (Plan Mode): Section 12 now describes five mandatory sub-sections (was four). New 12a — "Problem / Solution" — is two one-sentence lines (**Problem**: … / **Solution**: …) that anchor why the plan exists. Remaining sub-sections reletter: 12b Overall Plan Confidence, 12c Problem (Before) / Solution (After) tables, 12d Scope, 12e KPI Dashboard.
  • Template .claude/rules/planning/plan-checklist.md: Section 12 opens with Problem/Solution one-liners; Quick Validation gets a new checkbox; bumped to v9.
  • Template .claude/hooks/plan-sections-gate.sh: soft-warns when **Problem**: or **Solution**: lines are missing from the plan file (non-blocking); version reference bumped to v9.

Rationale

Every reviewer should see WHY a plan exists before reading confidence, KPIs, or per-fix tables. Section 0 preserves the raw user prompt; Section 12's new Problem/Solution distills it into one sentence each, in user-outcome terms. Writing Problem first catches scope creep: if either line spills to a second sentence, the plan is too broad or the intent isn't clear yet.

[5.0.1] - 2026-04-21

Patch release: Section 0 (verbatim user prompt) added as first mandatory section across plan-mode docs, template rules, and the ExitPlanMode hook.

Changed

  • Chapter 01 (Plan Mode): now describes 15 mandatory sections (was 14). New Section 0 — "Original User Prompt (verbatim, preserved)" — is the immutable, first section of every plan. Hook-enforced as blocking: ExitPlanMode refuses to submit without it, and soft-warns if the heading exists but no blockquote body is present.
  • Chapter 03 (Brainstorming): updated cross-reference from "14 sections" to "15 sections (starting with the verbatim user prompt)".
  • Template .claude/rules/planning/plan-checklist.md: added Section 0 template + Quick Validation checkbox; bumped to v8.
  • Template .claude/rules/planning/plan-link.md: header template now includes Section 0 blockquote immediately after metadata.
  • Template .claude/hooks/plan-sections-gate.sh: detects Section 0 heading + blockquote content; MISSING list now /12; version reference bumped to v8.

Rationale

Plans drift across revisions. The most common failure mode wasn't wrong tests or missing observability — it was paraphrased prompts. Freezing the original user prompt at the top gives every reviewer a reference point for detecting drift.

[5.0.0] - 2026-04-21

Major release: the guide is reorganized around 6 topical Parts, installation is now manifest-driven with three tiers, and compatibility is refreshed for Claude Code 2.1.111+.

Added

New structure: 6 topical Parts

  • Part I — Foundation (5 chapters)
  • Part II — Workflow (6 chapters)
  • Part III — Extension (9 chapters)
  • Part IV — Context Engineering (7 chapters)
  • Part V — Advanced (7 chapters)
  • Part VI — Reference (6 chapters)

New install tiers in best-practices/manifest.json

  • Core: 8 rules, 3 skills, 1 hook — newcomer-friendly
  • Recommended: +22 rules, +13 skills, +6 hooks — working developer
  • Full: +34 rules, +28 skills, +5 hooks + governance scaffolding — power user

New chapters (either brand new or synthesizing across multiple old chapters):

  • Part I: 01 Installation, 02 CLAUDE.md primer, 03 Project structure, 04 First session, 05 Setup troubleshooting
  • Part II: 02 TDD, 03 Brainstorming, 06 Commit and PR
  • Part III: 03b Claude Agent SDK (ported+modernized from old ch.43), 07 Slash commands (commands→skills migration reference)
  • Part IV: 06 Context Governance (7-layer system), 07 Skill lifecycle
  • Part V: 01 AI DNA shared-layer, 07 Session-end and defrag
  • Part VI: 01 CC version history (collapses 9 ship-logs), 02 CLI flags + env, 03 Hook event catalog (27 events), 04 Skill catalog, 05 MCP server catalog

Other additions

  • jekyll-redirect-from plugin for URL stability across the restructure
  • best-practices/manifest.json as authoritative tier definition
  • best-practices/test-manifest-resolve.sh self-test (12 assertions)
  • ROADMAP-v5.md public roadmap
  • docs/guide/_redirect-plan.md internal redirect mapping (51 entries)

Changed

  • install.sh rewritten to be manifest-driven (replaces hardcoded file lists from v4.x)
  • New install flags: --recommended, --full, --dry-run
  • CC compatibility updated from 2.1.99 to 2.1.111+
  • README hero metrics restated with honest counts
  • All chapters verified against current CC flags, hooks (27 events), skills, MCP patterns

Removed

Hardcoded installer flags (v4.x legacy):

  • --commands (commands migrated to skills at CC 2.1.88)
  • --all-rules (folded into --recommended)
  • --with-hooks (folded into tier composition)

Pruned chapters (content obsolete or superseded):

  • 9 version ship-logs (54, 57, 60, 64, 66, 70, 71, 72, 73) → consolidated into part6/01
  • Pre-native-skill-loading content (05, 30, 30b)
  • Branch-context chapters (31, 31b, 33) — replaced by path-scoped rules
  • Scattered orchestration/config chapters (41, 46, 47, 48, 49) — absorbed into relevant Parts
  • UI/UX rules (52) — out of CC guide scope
  • Research logs (55, 56, 59, 67, 68, 69)

Absorbed + deleted with redirects:

  • 02, 04, 06, 12, 13, 14, 15, 18, 19, 19b, 22, 23, 25, 26, 27, 28, 29, 32, 34, 35, 36, 37, 38, 39, 40, 42, 43, 44, 45, 50, 51, 53, 58, 61, 62, 63, 65, 74, 75, 76, 77, 78

Fixed

  • install.sh comments now honestly state tier counts (old comments claimed "35 rules" while installing 19)
  • CC version references updated across all chapters (no more "CC 2.1.99 compatible" on chapters describing 2.1.111 features)
  • Hero metrics no longer cite unsourced performance claims
  • settings.json mcpServers myth debunked — MCP servers register via claude mcp add stored in ~/.claude.json

[4.4.0] - 2026-04-20

Added

  • Chapter 78: Self-Telemetry for Claude Code — hooks + jsonl pattern for measuring your own CC usage (tool calls, subagent dispatches, skill invocations, session KPIs) without OTEL. First-published SubagentStart / SubagentStop stdin payload schema for CC 2.1.111 (validated by live probe). Documents the FIFO correlation pattern for duration measurement without tool_use_id, the grep -c ... || echo 0 double-output gotcha + _safe_count fix, and the event: discriminator convention for multi-event jsonl streams. Covers what is NOT measurable via hooks (OTEL file exporter unsupported, PermissionGranted event absent, no claude metrics CLI).

Changed

  • Ch.13 hooks: harmonized the "Use Absolute Paths for Hook Commands" example to use $CLAUDE_PROJECT_DIR/.claude/hooks/... instead of a hardcoded /home/user/project/... path, aligning with the $CLAUDE_PROJECT_DIR section immediately above.

[4.3.0] - 2026-04-12

Added

  • Chapter 71: Claude Code 2.1.93-2.1.94 features (Mantle on Bedrock, default effort high, plugin skill YAML hooks, Slack MCP)
  • Chapter 72: Claude Code 2.1.95-2.1.97 features (Focus view Ctrl+O, statusline refreshInterval, workspace.git_worktree, Accept Edits safe-env, /agents running indicator)
  • Chapter 73: Claude Code 2.1.98-2.1.99 features (Monitor tool, settings resilience, 6 permission bypass fixes, subagent MCP inheritance, worktree Read/Edit, PID namespace sandbox, OTEL TRACEPARENT, /agents tabbed UI, /team-onboarding, OS CA cert trust)
  • Chapter 74: Claude Code Monitor Tool — standalone reference with ScheduleWakeup decision matrix
  • Chapter 75: Claude Code Statusline Patterns — workspace.git_worktree, refreshInterval, 3 example scripts
  • Chapter 70: Claude Code 2.1.89-2.1.92 features (bundled from WIP)
  • Chapter 69: Knowledge Harvest Adoption (bundled from WIP)
  • 3 new best-practices rules: concurrency-partitioning, fail-closed-defaults, source-validation (bundled from WIP)

Changed

  • Ch.13 hooks: appended 2026-04 Updates section (settings resilience, permissions.deny precedence, hook path fix, 27+ events)
  • README: "25 hook events" → "27+ hook events", "60 chapters" → "65 chapters", "26 rules" → "29 rules", added "CC 2.1.99 compatible"
  • CITATION.cff: bumped to v1.3.0, updated date-released, synced abstract counters (27+ hooks, 65 chapters, 29 rules)
  • llms.txt: added chapters 69-75
  • docs/index.md: updated TOC and stats
  • 14 chapters refreshed with minor updates (bundled from WIP)

[4.2.0] - 2026-04-01

Added

  • Chapter 69: Knowledge Harvest Adoption — validation protocol for community articles (3-step: CLI check, guide search, reject unvalidated), progressive disclosure splits (80%+ body reduction), 2 new skills (/iterative-review, /fix-ci), spec-first TDD gate
  • Chapter 58: Added "Progressive Disclosure Split" cookbook recipe (Section 6) — when to split skills over 300 lines, three-level loading cost, before/after line counts

[4.1.0] - 2026-03-28

Added

  • Chapter 64: Claude Code 2.1.84-2.1.86 features (conditional if hooks, skill description 250-char cap, Read compact format, paths YAML lists, idle-return prompt, system-prompt caching with ToolSearch, code intelligence plugins, verification hook pattern)
  • Verification hook template: task-verification-gate.sh (advisory PreToolUse hook for TaskUpdate)
  • paths: frontmatter added to 9 domain-specific best-practices rules for conditional loading

Changed

  • Best practices rules: Added paths: frontmatter to 9 domain-specific rules (api-key-hygiene, bash-filename-iteration, doctor-command, feature-toggle-principle, frontmatter-format, no-hardcoded-urls, primacy-recency, retry-circuit-breaker, test-preflight)
  • Template hooks: Updated post-tool-use example with if conditional field
  • Best practices version: 2.0.0 → 2.1.0

[4.0.0] - 2026-03-25

Added

  • Chapter 60: Claude Code 2.1.82-2.1.83 features (CwdChanged/FileChanged hooks, TaskOutput deprecation, MEMORY.md cap, transcript search, env scrub, managed-settings.d)
  • Chapter 61: Stack Audit & Maintenance Patterns (frontmatter compliance, /document v3 three-level analysis, /retrospective skill pipeline, four-area audit methodology)
  • 11 new universal rules in best-practices: frontmatter-format, diagnostic-first, no-band-aids, autonomous-fixing, task-tracking, retry-circuit-breaker, test-preflight, feature-toggle-principle, api-key-hygiene, no-hardcoded-urls, bash-filename-iteration
  • Debugging rules category (diagnostic-first, no-band-aids) — prevents patch-on-patch cycles
  • 6 new BEST-PRACTICES.md sections (18-23): Frontmatter Format, Debugging Methodology, Task Tracking, External API Safety, Environment Hygiene, Test Preflight

Changed

  • Template commands: Fixed frontmatter format in all 7 commands (allowed_tools → allowed-tools, JSON arrays → comma-separated)
  • Best practices rules: 12 → 23 universal rules
  • Template rules: 24 → 35 rules across debugging/, process/, technical/ categories
  • Session protocol rule: Updated for Claude Code 2.1.83 features
  • Plan checklist rule: Updated to 13 mandatory sections with KPI dashboard format
  • BEST-PRACTICES.md: 17 → 23 sections
  • docs/index.md: Updated stats (2.1.83, 21 hooks, 23 rules, 240+ patterns)
  • Best practices version: 1.2.0 → 2.0.0

Removed

  • Archived 7 deprecated pre-prompt hook chapters to docs/archive/ (16, 17, 20, 21, 24, 29b, pre-prompt-hook-complete-guide)
  • Pre-prompt hook skill matching system fully superseded by Claude Code native skill loading (v2.1.76+)

[3.5.0] - 2026-03-06

Changed

  • Updated hook events from 18 to 19 (added InstructionsLoaded)
  • Updated TeammateIdle and TaskCompleted hooks to document {"continue": false, "stopReason": "..."} JSON output support for stopping teammates
  • Updated model documentation with effort levels (low/medium/high) and "ultrathink" keyword for high effort

Added

  • Chapter 13: InstructionsLoaded hook event (fires when CLAUDE.md or .claude/rules/*.md files are loaded), agent_id/agent_type fields in hook event data, worktree field in status line hooks
  • Chapter 38: Effort levels section (Opus 4.6 defaults to medium effort), includeGitInstructions and sandbox.enableWeakerNetworkIsolation settings
  • Skill Activation System: ${CLAUDE_SKILL_DIR} variable documentation, description field requirement clarification, colon-in-description fix note

[3.4.0] - 2026-03-03

Added

  • Guide #53: Pre-Validation Probe — A 2-minute practice that catches fundamentally wrong plans before implementation begins. Every plan rests on assumptions about reality; the probe verifies those assumptions with real evidence (grep, curl, file reads, test runs, screenshots — any tool counts) during plan mode, before approving. Includes assumption test table, feasibility checks, go/no-go verdict, examples across 12 project types (frontend, backend, performance, security, CI/CD, etc.), anti-patterns, and real-world wins. Integrates as Section 0.1 in the Plan Mode Checklist.
  • Template: ~/.claude/rules/planning/pre-validation.md — adoptable rule file for pre-validation probe enforcement

Changed

  • Guide #45: Plan Mode Checklist updated from 12 to 13 mandatory sections (added Section 0.1: Pre-Validation Probe as a blocking gate between TL;DR and Requirements)
  • Template: plan-checklist.md updated with Section 0.1 and Quick Validation checklist additions
  • Best Practices: validation-workflow.md updated from 5 to 6 Pre-Implementation Gates (added Gate 0: Pre-Validation Probe)

[3.3.0] - 2026-02-24

Changed

  • Updated hook events from 14 to 18 (added Setup, ConfigChange, WorktreeCreate, WorktreeRemove)
  • Updated installation instructions: npm deprecated, now uses claude install / curl installer
  • Fixed PreToolUse "approve" → "allow" in example settings
  • Added prompt-based and agent-based hook types to Chapter 13
  • Added 1M context window documentation across guide
  • Updated model references to Opus 4.6, Sonnet 4.6, Haiku 4.5

Added

  • Chapter 13: 4 new hook events, prompt/agent hook types, additionalContext, frontmatter hooks
  • Chapter 36: Worktree isolation, claude agents CLI
  • Chapter 38: 1M context window, fast mode, updated pricing
  • Chapter 4: Native TaskCreate/TaskUpdate/TaskList/TaskGet documentation
  • Chapter 45: plansDirectory configuration
  • Chapter 46: Permission precedence (ask overrides allow)
  • README/index: Updated metrics, installation, SEO keywords

[3.2.0] - 2026-02-24

Added

  • Guide #13: File Size / Modularity Enforcement Hooks -- new production example showing how to pair PreToolUse + PostToolUse hooks for code quality enforcement. PreToolUse catches new oversized files before they hit disk (reads tool_input.content line count). PostToolUse detects growth in existing files using a /tmp cache with md5sum keys (only warns on >500L files if they grew by 20+ lines, avoiding noise on legacy violations). Both hooks are non-blocking (exit 0 always). Includes complete scripts, settings.json config, design decisions table, and practical usage examples.
  • Examples: Added file-size-precheck.sh and file-size-warning.sh to examples/production-claude-hooks/hooks/

[3.1.0] - 2026-02-23

Added

  • Guide #51: Persistent Memory Patterns — Inspired by claude-mem (30K+ stars). Four patterns that automatically persist session knowledge to Basic Memory MCP with zero manual effort: (1) Auto-Observation PostToolUse hook — selective JSONL capture of file edits and significant Bash commands, async <50ms, no AI processing; (2) Auto Session Summary SessionEnd hook — reads git log (2h window) + JSONL observations → writes structured Basic Memory note; (3) Progressive Disclosure Search — 3-layer workflow (search IDs → search_notes preview → fetch full) with 10x token savings vs loading all results; (4) Observation Taxonomy — 6 standardized types ([bugfix]/[feature]/[refactor]/[change]/[discovery]/[decision]) + 5 concepts (#how-it-works/#problem-solution/#gotcha/#pattern/#trade-off). All patterns are global (~/.claude/) not project-specific. Includes design comparison: hook approach vs claude-mem worker service.

[3.0.0] - 2026-02-23

Added

  • Guide #50: Verification Feedback Loop - Boris Cherny's insight that giving Claude a way to verify its work 2-3x the quality. Three components: (1) /verify command with $! dynamic context injection for zero-round-trip verification, (2) verify-app agent with 3-tier verification (static/health/tests) and tech stack auto-detection, (3) Stop hook that detects source code changes and nudges verification. Includes global vs project scope explanation and adoptable templates.

[2.9.0] - 2026-02-23

Added

  • Guide #49: Workflow Resilience Patterns - four production-tested patterns for handling common session failure modes. (1) Autonomous Fixing: fix/ask decision framework for when Claude breaks something vs when to ask the user. (2) Correction Capture: persist user corrections to Basic Memory MCP or .claude/rules/ so they're never repeated. (3) Task Tracking Conventions: when and how to use TaskCreate/TaskUpdate with 6 hygiene rules. (4) Sideways Detection: 3-strike rule, scope creep, wrong assumptions, and blocked triggers for mid-execution re-planning. Each pattern includes an adoptable rule template.

[2.8.0] - 2026-02-19

Added

  • Guide #48: Lean Orchestrator Pattern - new chapter on defeating context rot during multi-task plan execution. Covers delegation decision matrix (when to spawn subagents vs work inline), parallel execution for independent tasks, context budget targets (<20% orchestrator), filesystem verification patterns, and common mistakes. Inspired by GSD project analysis. Includes adoptable rule template.

[2.7.0] - 2026-02-18

Changed

  • Guide #45: Plan Mode Quality Checklist expanded from 10 to 11 mandatory sections. Added Section 10 (Modularity Enforcement - blocking gate with 4 sub-checks: File Size Gate, Layer Separation Gate, Extraction Gate, God File Prevention). Updated template, Chapter 47 references, and design decisions.
  • Guide #26: Added "Global vs Project Rule Deduplication" section. Best practice for preventing double-loading when identical rules exist in both ~/.claude/rules/ and .claude/rules/. Includes audit workflow and token savings evidence (~1,139 lines saved).

[2.6.0] - 2026-02-17

Changed

  • Guide #18: Expanded from "Perplexity Cost Optimization" to "MCP Cost Control" with hook enforcement. Added PreToolUse/PostToolUse two-hook sandwich pattern, global settings placement (user-level), generalizable MCP cost control table, and updated validation checklist.

[2.5.0] - 2026-02-16

Changed

  • Guide #45: Plan Mode Quality Checklist expanded from 8 to 10 mandatory sections. Added Section 8 (File Change Summary - table of affected files with action and description) and Section 9 (Plan Summary TL;DR - 3-5 bullet points). Added Plan File Metadata section (branch, timestamp, topic, keywords for discoverability). Updated design decisions and takeaways.
  • index.md: Fixed stale metrics (Skills 162+ → 226+, Chapters 37+ → 42, Agent Patterns 3 → 5). Added Planning & Quality navigation section. Added missing Advanced Topics links (chapters 38-44).

[2.4.0] - 2026-02-14

Added

  • Guide #45: Plan Mode Quality Checklist - 8 mandatory plan sections (requirements clarification, existing code check, over-engineering prevention, best practices, modular architecture, documentation, E2E testing, observability). Two complementary approaches: rules file (passive, always in context) and user-invocable skill (on-demand). Covers the limitation of no plan mode hook event.

[2.3.0] - 2026-02-13

Added

  • Guide #40: Agent Orchestration Patterns - 5 core workflow architectures (Chain, Parallel, Routing, Orchestrator-Workers, Evaluator-Optimizer), query classification, subagent budgeting
  • Guide #41: Evaluation Patterns - Anthropic's 6 eval best practices (capability vs regression, outcome-based grading, pass^k consistency, LLM-as-judge, transcript review, saturation monitoring)
  • Guide #42: Session Memory & Compaction - SESSION_MEMORY_PROMPT template, PreCompact hooks, 75% rule, recovery patterns
  • Guide #43: Claude Agent SDK - Stateless vs stateful agents, tool permissions (allowed_tools vs disallowed_tools), MCP integration, plan mode
  • Guide #44: Skill Design Principles - Degrees of freedom framework, progressive disclosure, scripts as black boxes, negative scope, anti-clutter rules

Changed

  • Guide #13: Added "Always Exit 0" best practice, Python hook patterns, settings.local.json configuration
  • Guide #36: Added tool permission models (allowed_tools vs disallowed_tools distinction), query classification for agent routing, "Fresh Eyes" QA pattern
  • Guide #38: Added "Context Window Is a Public Good" principle, three-level progressive disclosure, scripts-as-black-boxes token savings

[2.2.0] - 2026-02-12

Changed

Removed

  • tools/trendradar-dashboard/ — moved to separate repository

[2.1.0] - 2026-02-12

Added

  • Guide #36: Task(agent_type) restriction patterns and examples for controlled sub-agent delegation
  • Guide #38: "When You MUST Override" section with real-world 213-skill measurement example

Changed

  • Guide #38: Budget measurement script updated to include project-level skills (.claude/skills/)

[2.0.0] - 2026-02-08

Added

  • SEO/AEO/GEO Overhaul - Complete search engine and AI discoverability optimization
  • robots.txt - AI crawler access (GPTBot, Claude-Web, PerplexityBot)
  • sitemap.xml - 35 pages for search engine indexing
  • _config.yml - Jekyll/GitHub Pages configuration with SEO tags
  • CITATION.cff - Academic and AI discovery metadata
  • docs/index.md - AEO-optimized landing page with FAQ structure
  • CONTRIBUTING.md - Community contribution guidelines
  • JSON-LD structured data for search engines

Changed

  • Repository renamed: claude-code-implementation-guide → claude-code-guide
  • README.md major overhaul with SEO keywords and hero section
  • All internal links updated to new repository URL

[1.5.0] - 2026-02-05

Added

  • Guide #35: Skill Optimization Maintenance - Long-term skill health patterns
  • Guide #34: Basic Memory MCP Integration - Persistent knowledge patterns
  • Guide #33: Branch-Specific Skill Curation - Per-branch skill optimization

Changed

  • STATUS.md updated with current metrics

[1.4.0] - 2026-01-20

Added

  • Guide #32: Document Automation - Auto-generation patterns
  • Guide #31: Branch-Aware Development - Multi-branch workflows
  • Guide #30: Blueprint Auto-Loading - Feature context patterns

[1.3.0] - 2026-01-10

Added

  • Guide #29: Branch Context System - JSON-based branch configuration
  • Guide #29: Comprehensive Skill Activation Testing - Validation methodology
  • Guide #28: Skill Optimization Patterns - Maintenance workflows

[1.2.0] - 2025-12-28

Added

  • Guide #27: Fast Cloud Run Deployment - GCP patterns
  • Guide #26: Claude Code Rules System - .claude/rules/ structure
  • Guide #25: Best Practices Reference - Anthropic-aligned patterns

[1.1.0] - 2025-12-20

Added

  • Guide #24: Skill Keyword Enhancement Methodology
  • Guide #23: Session Documentation Skill
  • Guide #22: wshobson Marketplace Integration
  • Guide #21: Pre-Prompt Optimization (370x improvement)
  • Guide #20: Skills Filtering Optimization (47-70% token savings)

Changed

  • Pre-prompt hook guide updated with 10k character limit patterns

[1.0.0] - 2025-12-14

Added

  • Initial release with core documentation
  • Quick Start Guide
  • Pre-Prompt Hook Complete Guide
  • Skill Activation System (88.2% accuracy)
  • MCP Integration Guide
  • Memory Bank Hierarchy
  • 19 detailed guide documents
  • Template CLAUDE.md
  • Skills library with 162+ skills
  • Example hooks and scripts

Version History Summary

Version Date Highlights
5.0.0 2026-04-21 6 topical Parts (~42 chapters), manifest-driven install (Core/Recommended/Full), CC 2.1.111+
4.4.0 2026-04-20 Ch.78 self-telemetry, SubagentStart/Stop payload schema, hook path harmonization
4.3.0 2026-04-12 CC 2.1.93-2.1.99 features, Monitor tool, statusline patterns, 3 new rules
4.2.0 2026-04-01 Knowledge harvest adoption: validation protocol, progressive disclosure splits, 2 new skills
3.3.0 2026-02-24 1M context, Opus 4.6/Sonnet 4.6/Haiku 4.5, 18 hooks, task mgmt, agents
3.2.0 2026-02-24 File size/modularity enforcement hooks (PreToolUse + PostToolUse paired)
3.1.0 2026-02-23 Persistent memory patterns (auto-observation, session summary, taxonomy)
3.0.0 2026-02-23 Verification feedback loop: /verify command, verify-app agent, Stop hook
2.9.0 2026-02-23 Workflow resilience: autonomous fixing, correction capture, tasks, sideways
2.8.0 2026-02-19 Lean orchestrator: context rot prevention via subagent delegation
2.7.0 2026-02-18 Plan mode: 11 sections + modularity gate + rule deduplication guide
2.6.0 2026-02-17 MCP cost control hooks (PreToolUse/PostToolUse sandwich pattern)
2.5.0 2026-02-16 Plan mode: 10 sections + file metadata (branch, timestamp, keywords)
2.4.0 2026-02-14 Plan mode quality checklist (8 mandatory sections)
2.3.0 2026-02-13 5 new chapters (orchestration, evals, compaction, SDK, skill design)
2.1.0 2026-02-12 Task restrictions, budget override
2.0.0 2026-02-08 SEO/AEO/GEO overhaul, repo rename
1.5.0 2026-02-05 Skill maintenance patterns
1.4.0 2026-01-20 Branch-aware development
1.3.0 2026-01-10 Context and testing systems
1.2.0 2025-12-28 Rules and best practices
1.1.0 2025-12-20 Pre-prompt 370x optimization
1.0.0 2025-12-14 Initial release