@izkac/forgekit 0.3.0 → 0.3.1

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.
Files changed (2) hide show
  1. package/README.md +83 -36
  2. package/package.json +4 -4
package/README.md CHANGED
@@ -1,36 +1,83 @@
1
- # @izkac/forgekit
2
-
3
- Portable agent skills + CLIs. One package, three bins:
4
-
5
- | Bin | Role |
6
- |-----|------|
7
- | **`forgekit`** | Install / list / update / uninstall skills × agents |
8
- | **`forge`** | Forge workflow sessions |
9
- | **`review`** | Thorough code review pipeline |
10
-
11
- ## Install
12
-
13
- ```bash
14
- npm i -g @izkac/forgekit
15
- forgekit install
16
- ```
17
-
18
- Interactive install uses arrow-key checkboxes — space to toggle, **`a` to select all**, enter to confirm. Pickers come pre-checked with what you already have installed; choosing the full set reconciles (new picks install, deselected ones are removed).
19
-
20
- Supported environments: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Windsurf, opencode — each into its global Agent-Skills directory.
21
-
22
- Non-interactive:
23
-
24
- ```bash
25
- forgekit install --skills forge,thorough-code-review --agents cursor,claude,copilot --force
26
- ```
27
-
28
- ## Docs
29
-
30
- - After install, full Forge reference: `~/.claude/skills/forge/docs/forge.md` (or the matching dir for your environment)
31
- - [How to use Forgekit](https://github.com/izkac/forgekit/blob/main/docs/usage.md)
32
- - [Repository README](https://github.com/izkac/forgekit#readme)
33
-
34
- ## License
35
-
36
- MIT
1
+ # @izkac/forgekit
2
+
3
+ **A disciplined development workflow for AI coding agents — portable across Claude Code, Cursor, Codex, Copilot, Gemini, Windsurf, and opencode.**
4
+
5
+ ![npm](https://img.shields.io/npm/v/@izkac/forgekit) ![node](https://img.shields.io/node/v/@izkac/forgekit) ![license](https://img.shields.io/npm/l/@izkac/forgekit)
6
+
7
+ AI agents are fast but undisciplined: they claim success on stubs, skip verification, over-build, and quietly leave the human behind. Forgekit installs a **gated workflow** that holds an agent to the standards a senior engineer would — and gives you, the operator, the visibility and control to stay in charge of a whole fleet of them.
8
+
9
+ One package, three command-line tools:
10
+
11
+ | Bin | What it does |
12
+ |-----|--------------|
13
+ | **`forgekit`** | Install / update the Forge skills into any supported agent |
14
+ | **`forge`** | Run the Forge workflow — sessions, integrity gates, fleet control |
15
+ | **`review`** | Standalone multi-pass thorough code review |
16
+
17
+ ## Why Forge (the workflow)
18
+
19
+ Forge moves every substantial change through explicit phases — **triage → brainstorm → plan → implement → verify → review → done** — and enforces them with **executable gates**, not honor-system promises:
20
+
21
+ - **Operator brief** — before any code is written, the agent produces a plain-language HTML explainer of what's about to be built (with diagrams). It opens in your browser and *gates implementation*: you approve what you actually understand, not a wall of spec markdown.
22
+ - **Runtime spine** — every change declares how each capability wires up: library → runtime owner → what it writes → the evidence it works. No orphaned code that's "done" but never called.
23
+ - **Executable product loop** — for jobs, workers, and pipelines the end-to-end acceptance is a *command that must exit green*, not a paragraph claiming it would.
24
+ - **Registered deferrals** — "wire it up later" is only allowed when tracked; the finish gate refuses unresolved ones.
25
+ - **Session scorecard** every session leaves a graded, measurable trail (A–F).
26
+
27
+ The result: agents that can't fake completion, and a workflow you can trust to run with less babysitting.
28
+
29
+ ## Fleet control — command every session from one terminal
30
+
31
+ Running agents across three IDEs and two terminals? `forge fleet` is a single control surface over all of them, across every project and engine:
32
+
33
+ ```bash
34
+ forge fleet list # every session: phase, task progress, engine, activity
35
+ forge fleet watch # live-refreshing dashboard
36
+ forge fleet view <s> # detail + live transcript tail (Claude Code)
37
+ forge fleet send <s> "pause and report" # message any session; --all broadcasts
38
+ ```
39
+
40
+ Sessions auto-register the moment they touch Forge — including ones running inside Claude Desktop or Cursor. You see where everything stands at a glance and steer any of them without switching windows.
41
+
42
+ ## Portable — one workflow, every agent
43
+
44
+ The same skills and the same discipline, installed into each agent's native skills directory:
45
+
46
+ **Claude Code · Cursor · Codex CLI · GitHub Copilot · Gemini CLI · Windsurf · opencode**
47
+
48
+ Planning is engine-flexible too: use the **OpenSpec** CLI if you have it, or the **built-in specs engine** (plain markdown, same layout) if you don't. Optional **ADRs** archive each finished change into an Architecture Decision Record.
49
+
50
+ ## Install
51
+
52
+ ```bash
53
+ npm i -g @izkac/forgekit
54
+ forgekit install # interactive: pick skills + agents (space toggles, `a` = all)
55
+ ```
56
+
57
+ Non-interactive:
58
+
59
+ ```bash
60
+ forgekit install --skills forge,thorough-code-review --agents claude,cursor --force
61
+ ```
62
+
63
+ Then wire a project and start working:
64
+
65
+ ```bash
66
+ cd your-project
67
+ forge init --claude --cursor # slash commands, rules, hooks, .forge/
68
+
69
+ # in the agent chat:
70
+ /forge add a health endpoint that returns { ok: true }
71
+ ```
72
+
73
+ The agent triages, brainstorms with you, plans a tracked change, writes the operator brief for your approval, then implements task-by-task with TDD and per-task review — refusing to call itself done until the gates are green.
74
+
75
+ ## Docs
76
+
77
+ - **[How to use Forgekit](https://github.com/izkac/forgekit/blob/main/docs/usage.md)** — full tutorial: install, project wiring, worked examples, integrity gates, fleet, cheat sheet
78
+ - **[Full Forge reference](https://github.com/izkac/forgekit/blob/main/skills/forge/docs/forge.md)** — phases, pace matrix, integrity rules (also installed at `~/.claude/skills/forge/docs/forge.md`)
79
+ - **[Repository](https://github.com/izkac/forgekit#readme)**
80
+
81
+ ## License
82
+
83
+ MIT. The Forge skill and several of its workflows are adapted from [Superpowers](https://github.com/obra/superpowers) (MIT).
package/package.json CHANGED
@@ -1,12 +1,12 @@
1
1
  {
2
2
  "name": "@izkac/forgekit",
3
- "version": "0.3.0",
3
+ "version": "0.3.1",
4
4
  "description": "Forgekit CLIs — forgekit (install), forge (workflow), review (code review)",
5
5
  "type": "module",
6
6
  "bin": {
7
- "forgekit": "./bin/forgekit.mjs",
8
- "forge": "./bin/forge.mjs",
9
- "review": "./bin/review.mjs"
7
+ "forgekit": "bin/forgekit.mjs",
8
+ "forge": "bin/forge.mjs",
9
+ "review": "bin/review.mjs"
10
10
  },
11
11
  "files": [
12
12
  "bin",