kodelyth-ecc 1.4.0 → 1.5.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/wiki/FAQ.md CHANGED
@@ -4,139 +4,225 @@
4
4
 
5
5
  ## General
6
6
 
7
- ### Is this really free?
7
+ ### What is Kodelyth ECC?
8
8
 
9
- Yes. MIT license. No paid tier, no API key required, no usage limits. The tool itself is free you pay only for whatever AI platform you use (Claude Code, Antigravity, etc.) through Anthropic or Google directly.
9
+ ECC (Enhanced Coding Companion) is a production-grade AI coding toolkit. It gives your AI coding tool 59 specialist agents, 188 skills, 80 slash commands, 18+ automation hooks, and a local self-learning memory system. Works with Claude Code, Windsurf, Cursor, Codex CLI, Antigravity, and OpenCode.
10
10
 
11
- ### Does it work with any language?
11
+ ### Is it free?
12
12
 
13
- Yes. The agents are language-agnostic. There are also language-specific rule bundles (TypeScript, Python, Go, Rust, Java, Kotlin, C++, Dart, PHP, Swift, Perl, web) that can be installed via profiles.
13
+ Yes. MIT license. No usage restrictions. No paid tier. No telemetry. No cloud dependency.
14
14
 
15
- ### How is this different from just using Claude Code alone?
15
+ ### Does it work with any AI model?
16
16
 
17
- Out of the box, Claude Code is a general-purpose assistant. Kodelyth ECC adds:
18
- - 53 specialist agents that are deeply opinionated in their domains
19
- - Automated hooks that run quality checks without you asking
20
- - Slash commands for repeatable workflows
21
- - Domain knowledge skills that stay active for the session
22
- - Always-on principles (no emoji, budget-first model selection, SVG over decoration)
17
+ Agents and skills are model-agnostic markdown files. They work with Claude, GPT, Gemini, Llama, and any other model your AI tool uses. The memory system is also model-agnostic.
18
+
19
+ ### Do I need to configure anything after install?
20
+
21
+ No. Hooks, rules, and memory injection activate automatically. You don't need to configure anything to get the core features.
23
22
 
24
23
  ---
25
24
 
26
25
  ## Installation
27
26
 
28
- ### Install fails with "permission denied"
27
+ ### What's the easiest way to install?
29
28
 
30
29
  ```bash
31
- chmod +x install.sh
32
- ./install.sh
30
+ npx kodelyth-ecc
33
31
  ```
34
32
 
35
- ### Install fails on Windows
36
-
37
- Use PowerShell (not Command Prompt):
38
-
39
- ```powershell
40
- Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
41
- .\install.ps1
42
- ```
33
+ Requires Node 18+. Runs on any OS.
43
34
 
44
- ### How do I update to a new version?
35
+ ### What install targets are available?
45
36
 
46
- ```bash
47
- cd kodelyth-ecc
48
- git pull
49
- ./install.sh
50
- ```
37
+ | Target | What it installs to |
38
+ |---|---|
39
+ | `claude-home` (default) | `~/.claude/` — Claude Code global |
40
+ | `antigravity` | `~/.config/google-gemini/` or Antigravity dir |
41
+ | `windsurf-project` | `.windsurf/` in current project |
42
+ | `windsurf-home` | `~/.codeium/windsurf/` global |
43
+ | `cursor-project` | `.cursor/` in current project |
44
+ | `codex-home` | `~/.codex/` |
45
+ | `opencode` | `~/.opencode/` |
51
46
 
52
- The install script is idempotent — safe to run multiple times.
47
+ ### What are install profiles?
53
48
 
54
- ### How do I uninstall?
49
+ Profiles install language-specific rule modules:
55
50
 
56
- Claude Code:
57
51
  ```bash
58
- rm -rf ~/.claude/agents ~/.claude/skills ~/.claude/hooks ~/.claude/commands ~/.claude/rules
52
+ npx kodelyth-ecc --profile nextjs # TypeScript rules
53
+ npx kodelyth-ecc --profile python-api # Python rules
54
+ npx kodelyth-ecc --profile fullstack # TypeScript + Python + Go
55
+ npx kodelyth-ecc --profile mobile # Kotlin + Swift
56
+ npx kodelyth-ecc --profile backend # Go + Python + Java
59
57
  ```
60
58
 
61
- Antigravity:
62
- ```bash
63
- rm -rf your-project/.agent/
64
- ```
59
+ ### Can I install to multiple targets?
60
+
61
+ Yes. Run the installer multiple times with different `--target` flags.
65
62
 
66
63
  ---
67
64
 
68
65
  ## Agents
69
66
 
70
- ### How do I use an agent?
67
+ ### Do I have to know agent names?
71
68
 
72
- ```
73
- use agent-name
74
- ```
69
+ No. The intent routing rule (`rules/common/agent-intent-routing.md`) analyzes every message you type and routes to the right agent automatically. You can also explicitly invoke agents:
75
70
 
76
- Example:
77
71
  ```
78
72
  use debug-detective
79
- use kodelyth-advisor
80
- use security-reviewer
73
+ @code-reviewer
74
+ invoke security-reviewer
81
75
  ```
82
76
 
83
- ### Which agent should I use first?
77
+ ### How do I know which agent handled my request?
84
78
 
85
- If you're not sure, always start with:
79
+ ECC always acknowledges routing explicitly:
86
80
 
87
81
  ```
88
- use kodelyth-advisor
82
+ Routing to debug-detective
83
+ Tip: type "use debug-detective" to invoke directly next time.
89
84
  ```
90
85
 
91
- It will assess your situation and direct you to the right agent or skill.
86
+ ### Can I add my own agents?
87
+
88
+ Yes. Create a markdown file with YAML frontmatter in `~/.claude/agents/`. See [CONTRIBUTING.md](https://github.com/sifxprime/kodelyth-ecc/blob/main/CONTRIBUTING.md) for the template.
89
+
90
+ ### What's the difference between planner and pair-programmer?
91
+
92
+ `planner` creates a full implementation plan (PRD, architecture, task list) before writing any code. `pair-programmer` is your pre-code thinking partner — it catches wrong approaches, missing edge cases, and premature optimizations before a line is written. Use `pair-programmer` first for critical decisions, then `planner` for the full breakdown.
93
+
94
+ ---
95
+
96
+ ## Memory System
97
+
98
+ ### What is the self-learning memory?
99
+
100
+ The memory system (v1.4.0+) captures problems and solutions from your past sessions and recalls them when relevant in future sessions. It runs locally at `~/.kodelyth/memory/` with zero cloud involvement.
101
+
102
+ ### How does memory capture work?
103
+
104
+ ECC never auto-stores anything. The Stop hook runs `extract.js` at session end, extracts learning candidates using heuristics, and writes them to `pending-review.jsonl`. You review and confirm:
105
+
106
+ ```bash
107
+ node scripts/memory/cli.js list # see what's pending
108
+ /memory review-pending # review in your AI session
109
+ /memory remember "title" --approach "..." --tags a,b
110
+ ```
111
+
112
+ ### What is auto-recall?
113
+
114
+ Auto-recall (v1.4.1) fires a `UserPromptSubmit` hook on every meaningful prompt. It runs BM25 search against your memory and injects relevant matches into the AI's context before it responds. No commands needed.
115
+
116
+ ### How does repeat suppression work?
117
+
118
+ Each session tracks which memory IDs were already surfaced (stored in `~/.kodelyth/memory/session-surfaced-<id>.json`). If the same memory was shown earlier in the session, it is not shown again. Different topics in the same session will still surface their relevant memories.
119
+
120
+ ### What prompts does auto-recall skip?
121
+
122
+ Auto-recall skips:
123
+ - Empty prompts
124
+ - Trivial one-word responses (`ok`, `yes`, `no`, `thanks`, `cool`, `done`)
125
+ - Agent/command invocations starting with `use`, `@`, `/`, or `invoke`
126
+ - Prompts under 12 characters
127
+ - Prompts with fewer than 2 meaningful tokens after stop-word removal
92
128
 
93
- ### Can I use multiple agents in one session?
129
+ ### Can I use memory without Claude Code?
94
130
 
95
- Yes. You can delegate to agents sequentially or in parallel. Example:
131
+ Manual memory (`/memory remember`, `/memory search`) works everywhere. Auto-recall via `UserPromptSubmit` hook only fires in Claude Code — other platforms don't currently expose a pre-prompt hook. Session-start injection works in all platforms.
96
132
 
133
+ ### Where is memory stored?
134
+
135
+ Default: `~/.kodelyth/memory/`
136
+
137
+ Override: `export KODELYTH_MEMORY_DIR=/your/path`
138
+
139
+ ### How do I search my memory?
140
+
141
+ ```bash
142
+ node scripts/memory/cli.js search "stripe webhook"
143
+ /memory recall stripe webhook
97
144
  ```
98
- use pair-programmer ← think through the approach first
99
- use tdd-guide ← write tests
100
- use code-reviewer ← review the implementation
101
- use security-reviewer ← check for vulnerabilities
145
+
146
+ ### How do I delete a memory?
147
+
148
+ ```bash
149
+ node scripts/memory/cli.js forget <memory-id>
102
150
  ```
103
151
 
152
+ Soft-delete — the entry is marked deleted but not removed from the log file.
153
+
104
154
  ---
105
155
 
106
- ## Google Antigravity
156
+ ## Hooks
107
157
 
108
- ### Do hooks work in Antigravity?
158
+ ### What hooks run automatically?
109
159
 
110
- Antigravity does not have a hook execution engine. Hook behavior is approximated via `.agent/rules/kodelyth-always-on.md` which auto-loads every session and enforces the same principles (test reminders, branch naming, model selection).
160
+ | Hook | When | What it does |
161
+ |---|---|---|
162
+ | `memory/inject-start.js` | Session start | Injects relevant memories into context |
163
+ | `memory/auto-recall.js` | Every prompt | Real-time BM25 memory check per message |
164
+ | `memory/capture-stop.js` | Session end | Extracts learning candidates to pending review |
165
+ | `hooks/test-reminder` | After editing code | Reminds you to write tests |
166
+ | `hooks/smart-suggest` | After response | Suggests the next logical agent |
167
+ | `hooks/branch-name-check` | Before `git checkout -b` | Enforces branch naming convention |
111
168
 
112
- ### Do I need to say anything to Antigravity to activate ECC?
169
+ ### How do I disable a hook?
113
170
 
114
- No. After install, `.agent/rules/kodelyth-always-on.md` is loaded automatically every session. ECC is always active.
171
+ ```bash
172
+ export ECC_DISABLED_HOOKS=kodelyth:smart-suggest,kodelyth:test-reminder
173
+ ```
115
174
 
116
- ### Which Antigravity model should I use for daily work?
175
+ ### Do hooks slow down my session?
117
176
 
118
- **Gemini 3.1 Pro (Low)** for everyday coding tasks. Switch to **Gemini 3.1 Pro (High)** for complex features or multi-file refactoring. Use **Claude Sonnet 4.6 Thinking** for agent-heavy or reasoning tasks.
177
+ No. All hooks are async with a 5-second timeout. They write to stderr (advisory only) and never block tool execution.
119
178
 
120
179
  ---
121
180
 
122
- ## Contributing
181
+ ## Platform Questions
123
182
 
124
- ### How do I contribute a new agent?
183
+ ### Does ECC work with Windsurf?
125
184
 
126
- See [CONTRIBUTING.md](https://github.com/sifxprime/kodelyth-ecc/blob/main/CONTRIBUTING.md).
185
+ Yes. Use `--target windsurf-project` or `--target windsurf-home`. The installer generates a `.windsurfrules` file that Cascade auto-loads.
127
186
 
128
- Agents must follow the Kodelyth Standard:
129
- - GOD-level persona (decade-experienced specialist)
130
- - Model-agnostic (no model field in frontmatter)
131
- - Production-grade examples
132
- - No emoji in output — SVG badges or plain text only
187
+ ### Does ECC work with Google Antigravity?
133
188
 
134
- ### How do I report a security issue?
189
+ Yes. Use `--target antigravity`. Rules are flattened for Antigravity's required format.
135
190
 
136
- Email **sifxprime@kodelyth.com** with subject `[SECURITY] kodelyth-ecc <description>`.
191
+ ### Does memory work with cloud-AI platforms like Antigravity?
137
192
 
138
- See [SECURITY.md](https://github.com/sifxprime/kodelyth-ecc/blob/main/SECURITY.md) for full policy.
193
+ Manual memory commands (`/memory remember`, `/memory search`) work everywhere. Auto-extract from past sessions requires local session files — cloud platforms that store sessions server-side don't support this. The `UserPromptSubmit` auto-recall hook also requires Claude Code currently.
139
194
 
140
195
  ---
141
196
 
142
- ## Built and maintained by [Kodelyth](https://github.com/sifxprime) — [@sifxprime](https://github.com/sifxprime)
197
+ ## Troubleshooting
198
+
199
+ ### My agent isn't responding to intent routing
200
+
201
+ Check that `rules/common/agent-intent-routing.md` is installed. Run:
202
+
203
+ ```bash
204
+ ls ~/.claude/rules/common/
205
+ ```
206
+
207
+ If missing, re-run `npx kodelyth-ecc`.
208
+
209
+ ### Memory isn't being injected
210
+
211
+ Check the `hooks/memory/inject-start.js` hook is installed and listed in `hooks.json`. Verify `KODELYTH_MEMORY_DIR` points to a directory with memory entries.
212
+
213
+ ### Tests are failing after install
214
+
215
+ Run the test suite to verify installation:
216
+
217
+ ```bash
218
+ cd ~/.claude # or your install directory
219
+ npm test
220
+ ```
221
+
222
+ ---
223
+
224
+ ## Contributing
225
+
226
+ See [CONTRIBUTING.md](https://github.com/sifxprime/kodelyth-ecc/blob/main/CONTRIBUTING.md) for agent, skill, command, and hook templates.
227
+
228
+ Branch naming: `feat/`, `fix/`, `docs/`, `refactor/`, `test/`, `chore/`
package/wiki/Home.md CHANGED
@@ -1,50 +1,113 @@
1
- # Kodelyth ECC Wiki
1
+ # Kodelyth ECC Wiki
2
2
 
3
- > Production-grade AI coding agent toolkit for Claude Code, Google Antigravity, Cursor, Codex, and OpenCode.
3
+ **Kodelyth Enhanced Coding Companion** — production-grade AI coding toolkit for Claude Code, Windsurf, Cursor, Codex CLI, Antigravity, and OpenCode.
4
4
 
5
- **Version:** 1.1.0 | **License:** MIT | **Author:** [Kodelyth / @sifxprime](https://github.com/sifxprime)
5
+ **Version:** 1.4.1 | **License:** MIT | **Author:** [Kodelyth / @sifxprime](https://github.com/sifxprime)
6
6
 
7
7
  ---
8
8
 
9
- ## Quick Navigation
9
+ ## What is Kodelyth ECC?
10
+
11
+ ECC is a collection of 59 battle-tested agents, 188 skills, 80 commands, 18+ hooks, and 16 rules that make any AI coding tool dramatically more capable. It installs into your AI tool's config directory and activates automatically on every session.
12
+
13
+ Three things make ECC genuinely different from other AI config setups:
10
14
 
11
- | Page | Description |
12
- |------|-------------|
13
- | [Installation Guide](Installation-Guide) | Install on any platform in 30 seconds |
14
- | [Agent Reference](Agent-Reference) | All 53 specialist agents explained |
15
- | [Skill Reference](Skill-Reference) | All 185 skills and how to load them |
16
- | [Hook Reference](Hook-Reference) | 18+ automated hooks that run in background |
17
- | [Platform Support](Platform-Support) | Claude Code, Antigravity, Cursor, Codex, OpenCode |
18
- | [FAQ](FAQ) | Common questions answered |
15
+ **1. God-Tier Intent Routing** — every message you type is analyzed across 10 priority tiers and 50+ trigger patterns. The right specialist activates automatically. You don't need to know agent names.
16
+
17
+ **2. Local Self-Learning Memory** ECC captures problems and solutions from your sessions and recalls them when relevant. Completely local. BM25 full-text search. Zero cloud. Zero telemetry.
18
+
19
+ **3. Auto Chat Detection** on every meaningful prompt, ECC automatically checks your memory for relevant past solutions and injects them into the AI's context before it responds. No commands needed. The AI just knows.
19
20
 
20
21
  ---
21
22
 
22
- ## What is Kodelyth ECC?
23
+ ## Quick Install
23
24
 
24
- Kodelyth ECC is a collection of 53 specialist AI agents, 185 knowledge skills, 79 slash commands, and 18+ automated hooks — all free, all open source, MIT licensed.
25
+ ```bash
26
+ # Any machine with Node 18+ (recommended)
27
+ npx kodelyth-ecc
25
28
 
26
- Instead of one generic AI assistant doing everything, you get a team:
29
+ # macOS / Linux
30
+ curl -fsSL https://raw.githubusercontent.com/sifxprime/kodelyth-ecc/main/install.sh | bash
27
31
 
28
- - A **debugger** that never guesses and traces every bug to its root cause
29
- - A **security reviewer** that catches OWASP Top 10 before you commit
30
- - A **UX reviewer** that enforces WCAG 2.1 AA without touching your design
31
- - A **pair programmer** that catches the wrong approach before you write a line
32
- - A **migration guide** with full playbooks for React, Next.js, Python, Node, TypeScript, Java
32
+ # Windows PowerShell
33
+ irm https://raw.githubusercontent.com/sifxprime/kodelyth-ecc/main/install.ps1 | iex
33
34
 
34
- All working together. All free.
35
+ # GitHub (no npm)
36
+ npx github:sifxprime/kodelyth-ecc
37
+ ```
35
38
 
36
39
  ---
37
40
 
38
- ## Install
41
+ ## What Gets Installed
39
42
 
40
- ```bash
41
- git clone https://github.com/sifxprime/kodelyth-ecc.git
42
- cd kodelyth-ecc
43
- ./install.sh
43
+ | Component | Count | Where |
44
+ |---|---|---|
45
+ | Agents | 59 | `~/.claude/agents/` |
46
+ | Skills | 188 | `~/.claude/skills/` |
47
+ | Commands | 80 | `~/.claude/commands/` |
48
+ | Hooks | 18+ | `~/.claude/hooks/` |
49
+ | Rules | 16 | `~/.claude/rules/` |
50
+ | Memory hooks | 3 | `~/.claude/hooks/memory/` |
51
+
52
+ ---
53
+
54
+ ## Intent Routing — How It Works
55
+
56
+ ```
57
+ You type: "there's a bug in the auth middleware"
58
+ ECC: → Routing to debug-detective
59
+
60
+ You type: "this test fails every third run"
61
+ ECC: → Routing to flake-hunter
62
+
63
+ You type: "I need to cut the v2.0 release"
64
+ ECC: → Routing to release-captain
65
+
66
+ You type: "add Stripe webhooks"
67
+ └─ Memory auto-recall fires → finds past solution
68
+ injects context before AI responds
44
69
  ```
45
70
 
46
- See [Installation Guide](Installation-Guide) for full options.
71
+ ---
72
+
73
+ ## Wiki Pages
74
+
75
+ | Page | What it covers |
76
+ |---|---|
77
+ | [Installation Guide](Installation-Guide) | All install methods, targets, profiles, and options |
78
+ | [Agent Reference](Agent-Reference) | All 59 agents — purpose, trigger patterns, when to use |
79
+ | [Skill Reference](Skill-Reference) | All major skills and how to invoke them |
80
+ | [Hook Reference](Hook-Reference) | All hooks — when they fire and what they do |
81
+ | [Platform Support](Platform-Support) | Per-platform capabilities, limitations, and configs |
82
+ | [FAQ](FAQ) | Common questions including memory system |
83
+
84
+ ---
85
+
86
+ ## Quick Decision Table
87
+
88
+ | Situation | Agent to use |
89
+ |---|---|
90
+ | Bug with stack trace | `debug-detective` |
91
+ | Bug with no error (silent fail) | `silent-failure-hunter` |
92
+ | Build or type error | `build-error-resolver` |
93
+ | Security concern | `security-reviewer` |
94
+ | Code review before commit | `code-reviewer` |
95
+ | Planning a new feature | `planner` or `pair-programmer` |
96
+ | API breaking change check | `api-guardian` |
97
+ | Frontend UX / accessibility | `ux-reviewer` |
98
+ | Flaky tests | `flake-hunter` |
99
+ | Cut a release | `release-captain` |
100
+ | Broken git state | `git-rescue` |
101
+ | "Works on my machine" | `env-debugger` |
102
+ | Dependency conflicts | `dependency-doctor` |
103
+ | Memory / past solutions | `kodelyth-memory` |
104
+ | Don't know which agent | `kodelyth-advisor` |
47
105
 
48
106
  ---
49
107
 
50
- Built with care. Powered by **Kodelyth**.
108
+ ## Links
109
+
110
+ - **Repository:** [github.com/sifxprime/kodelyth-ecc](https://github.com/sifxprime/kodelyth-ecc)
111
+ - **npm:** [npmjs.com/package/kodelyth-ecc](https://www.npmjs.com/package/kodelyth-ecc)
112
+ - **Changelog:** [CHANGELOG.md](https://github.com/sifxprime/kodelyth-ecc/blob/main/CHANGELOG.md)
113
+ - **License:** MIT — free forever