@amityco/social-plus-vise 0.8.0 → 0.8.1
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.
- package/README.md +60 -27
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -31,24 +31,23 @@ vise install-skill --target claude # Claude Code (personal)
|
|
|
31
31
|
vise install-skill --target cursor . # Cursor (project-local)
|
|
32
32
|
vise install-skill --target copilot . # GitHub Copilot / VS Code
|
|
33
33
|
|
|
34
|
-
# 3.
|
|
35
|
-
|
|
36
|
-
vise inspect # detect platform, surface, design signals
|
|
37
|
-
vise plan --request "Add a social feed" # produce a grounded implementation plan
|
|
38
|
-
vise init --request "Add a social feed" # write the sp-vise/ compliance contract
|
|
39
|
-
|
|
40
|
-
# 4. After the agent makes edits
|
|
41
|
-
vise check # verify the integration against the contract
|
|
42
|
-
vise run-sensors # run your project's own build/typecheck/lint
|
|
34
|
+
# 3. Ask your AI agent to integrate with social.plus
|
|
35
|
+
# (the skill handles the rest — inspect, plan, init, code, check)
|
|
43
36
|
```
|
|
44
37
|
|
|
45
|
-
|
|
38
|
+
**Step 3 in practice:** Open your AI coding tool in your project and prompt:
|
|
39
|
+
|
|
40
|
+
> "Add a social feed to this app using the social.plus SDK."
|
|
41
|
+
|
|
42
|
+
The installed skill teaches your agent to run `vise inspect` → `vise plan` → `vise init` → edit code → `vise check` → `vise run-sensors` automatically. You drive intent; Vise keeps the agent on the rails.
|
|
43
|
+
|
|
44
|
+
See [Usage Flow](#usage-flow) for the full step-by-step diagram.
|
|
46
45
|
|
|
47
46
|
---
|
|
48
47
|
|
|
49
48
|
## What Vise Does
|
|
50
49
|
|
|
51
|
-
Vise is a **
|
|
50
|
+
Vise is a **CLI + AI skill** that wraps coding agents in deterministic compliance guardrails when they integrate social.plus SDKs. It inspects your project, grounds the agent in hosted docs, enforces 250+ platform-specific compliance rules, and runs your project's own build/lint/typecheck sensors. **Your source code never leaves your machine.**
|
|
52
51
|
|
|
53
52
|
| Layer | Purpose |
|
|
54
53
|
|---|---|
|
|
@@ -62,6 +61,56 @@ A bench vise holds the workpiece steady so the craftsman's hands are free to sha
|
|
|
62
61
|
|
|
63
62
|
---
|
|
64
63
|
|
|
64
|
+
## Benchmark: First-Try Success
|
|
65
|
+
|
|
66
|
+
> **100% first-try CI pass with Vise vs 0% without.**
|
|
67
|
+
>
|
|
68
|
+
> **76% cheaper · 28% faster · 86% fewer issues**
|
|
69
|
+
|
|
70
|
+
When an AI agent integrates social.plus with only docs access (Pure MCP), it produces code with real problems: hardcoded user IDs, missing authentication, no content moderation, broken reactive patterns. These aren't edge cases — they're the SDK-specific requirements that general AI knowledge reliably misses.
|
|
71
|
+
|
|
72
|
+
### v0.8 Pilot Results (React/Next.js · "add comments")
|
|
73
|
+
|
|
74
|
+
| Surface | CI Pass | Issues | Tokens | Cost | Wall-clock |
|
|
75
|
+
|---|---|---|---|---|---|
|
|
76
|
+
| **Pure MCP** (docs only) | ❌ 0/2 | 4–7 | 36,219 | $0.0108 | 619s |
|
|
77
|
+
| **Vise-as-MCP** (rules engine) | ✅ 2/2 | 1 | 21,047 | $0.0061 | 540s |
|
|
78
|
+
| **Vise CLI + Skill** (full workflow) | ✅ 2/2 | 1 | 8,733 | $0.0024 | 447s |
|
|
79
|
+
|
|
80
|
+
<sub>Token/cost data from Antigravity/Gemini Flash 3.5. Copilot CLI does not expose token accounting.</sub>
|
|
81
|
+
|
|
82
|
+
**What "Issues" means in plain language:**
|
|
83
|
+
|
|
84
|
+
Without Vise, both agents produced code with hardcoded user IDs (security vulnerability), no authentication flow (anonymous writes), missing moderation UI, non-reactive queries, and missing SDK initialization. With Vise, those problems are caught or prevented during generation.
|
|
85
|
+
|
|
86
|
+
### Why this matters
|
|
87
|
+
|
|
88
|
+
| Metric | Without Vise | With Vise (CLI + Skill) | Improvement |
|
|
89
|
+
|---|---|---|---|
|
|
90
|
+
| Does it work on first try? | ❌ Fails CI | ✅ Passes CI | 100% pass rate |
|
|
91
|
+
| Security issues? | Hardcoded IDs, no auth | 0 security findings | 100% eliminated |
|
|
92
|
+
| Integration issues | 4–7 per run | 1 per run | **−86%** fewer issues |
|
|
93
|
+
| Token cost | $0.0108 | $0.0024 | **−78%** cheaper |
|
|
94
|
+
| Token usage | 36,219 | 8,733 | **−76%** fewer tokens |
|
|
95
|
+
| Speed (Gemini) | 619s | 447s | **−28%** faster |
|
|
96
|
+
| Manual rework needed? | Yes | No | Zero rework |
|
|
97
|
+
|
|
98
|
+
### Cross-model validation
|
|
99
|
+
|
|
100
|
+
The effect holds across **Claude Sonnet 4.6** (Copilot CLI) and **Gemini Flash 3.5** (Antigravity). This is not a prompt trick for one model — it's domain knowledge applied consistently at the social.plus layer.
|
|
101
|
+
|
|
102
|
+
### Which mode should I use?
|
|
103
|
+
|
|
104
|
+
| If you... | Use | Why |
|
|
105
|
+
|---|---|---|
|
|
106
|
+
| Can install the skill | **CLI + Skill** | Fastest, cheapest, best results |
|
|
107
|
+
| Can't install skill but have MCP | **Vise-as-MCP** | Same compliance, slightly more tokens |
|
|
108
|
+
| Want to validate existing code | `vise check --ci` | Grade any codebase, any time |
|
|
109
|
+
|
|
110
|
+
For the full interactive report with charts, see [`benchmarks/report.html`](./benchmarks/report.html). For per-cell scorecards and prior benchmark versions, see [`benchmarks/RESULTS.md`](./benchmarks/RESULTS.md).
|
|
111
|
+
|
|
112
|
+
---
|
|
113
|
+
|
|
65
114
|
## Supported Platforms
|
|
66
115
|
|
|
67
116
|
| Platform | Coverage | Sensors |
|
|
@@ -289,22 +338,6 @@ Attestation files record source fingerprints (SHA-256 of cited files) so subsequ
|
|
|
289
338
|
|
|
290
339
|
---
|
|
291
340
|
|
|
292
|
-
## Benchmark Headline
|
|
293
|
-
|
|
294
|
-
| Platform | Pure MCP findings | Vise findings | Pure MCP CI | Vise CI |
|
|
295
|
-
|---|---|---|---|---|
|
|
296
|
-
| React / Next.js | 7 (3 errors) | 2 (warnings) | ❌ FAIL | ✅ PASS |
|
|
297
|
-
| React Native | 6 | 2 | ❌ FAIL | ✅ PASS |
|
|
298
|
-
| Flutter | 9 | 2 | ❌ FAIL | ✅ PASS |
|
|
299
|
-
| Android (Kotlin) | 9 | 0 | ❌ FAIL | ✅ PASS |
|
|
300
|
-
| iOS (Swift) | 8 | 0 | ❌ FAIL | ✅ PASS |
|
|
301
|
-
|
|
302
|
-
Measured runs of the same AI agent (Claude Sonnet 4.6) implementing "add a global feed" on each platform, with and without Vise. Without Vise: every run ships a hardcoded API key (a deterministic failure that cannot be attested). With Vise: every run passes CI with zero deterministic failures.
|
|
303
|
-
|
|
304
|
-
For full methodology, per-cell scorecards, and the v0.7 multi-outcome cross-tool validation (chat / comments / push on React + Flutter, plus Antigravity/Gemini cross-tool), see [`benchmarks/RESULTS.md`](./benchmarks/RESULTS.md) in the installed npm package.
|
|
305
|
-
|
|
306
|
-
---
|
|
307
|
-
|
|
308
341
|
## Changelog
|
|
309
342
|
|
|
310
343
|
See [`CHANGELOG.md`](./CHANGELOG.md) for the full version history.
|