getprismo 0.1.3 → 0.1.4

package/README.md

@@ -1,6 +1,6 @@
 # PrismoDev
 
-Prismo helps developers find and reduce token waste in AI coding workflows. It runs locally, supports Codex and Claude Code, and does not require API keys.
+PrismoDev finds and monitors token waste in local AI coding workflows, while Prismo core tracks exact API usage through the proxy.
 
 ```bash
 npx getprismo scan --usage
@@ -16,6 +16,16 @@ npx getprismo watch
 - Recommends safer `.claudeignore`, `CLAUDE.md`, and `AGENTS.md` patterns.
 - Works offline and does not connect to OpenAI, Anthropic, Cursor, or billing accounts.
 
+## Tracking Modes
+
+```text
+Local scan: no API keys, heuristic repo/context risk
+Local logs: exact when Codex/Claude logs expose token fields
+Prismo proxy: exact usage/cost when traffic uses Prismo base URL
+```
+
+PrismoDev does not claim exact billing for hidden subscription coding-agent sessions. When provider traffic does not flow through Prismo, local usage is based on available session logs and deterministic estimates.
+
 ## Quick Start
 
 ```bash
@@ -30,6 +40,8 @@ Use this flow to see value immediately:
 2. `setup` shows which tracking modes are possible, including Prismo proxy readiness.
 3. `watch --once` shows the current live local session view with warnings and next action.
 
+Testing PrismoDev with other developers? Use the [PrismoDev user testing playbook](docs/prismodev-user-testing.md).
+
 For a guided scan + context generation flow, run:
 
 ```bash

package/docs/prismodev-user-testing.md

@@ -0,0 +1,157 @@
+# PrismoDev User Testing Playbook
+
+Use this playbook to validate the CLI-first PrismoDev wedge with developers who use Codex, Claude Code, Cursor, or similar AI coding tools.
+
+## Goal
+
+Prove whether developers understand and value local AI coding-workflow visibility before building dashboard persistence, hooks, MCP interception, or live compression.
+
+Core positioning:
+
+> PrismoDev finds and monitors token waste in local AI coding workflows, while Prismo core tracks exact API usage through the proxy.
+
+## Who To Test With
+
+Recruit 5-10 developers who actively use at least one of:
+
+- Codex
+- Claude Code
+- Cursor
+- OpenAI-compatible coding tools
+- AI-heavy terminal workflows
+
+Prefer developers with real repos, long sessions, local logs, generated files, large test output, or MCP/tooling configs.
+
+## Exact Commands To Send
+
+Ask each tester to run these from a real project root:
+
+```bash
+npx getprismo@0.1.3 scan --usage --no-report
+npx getprismo@0.1.3 setup
+npx getprismo@0.1.3 watch --once
+```
+
+Optional follow-up if they find useful issues:
+
+```bash
+npx getprismo@0.1.3 scan --fix
+npx getprismo@0.1.3 optimize
+npx getprismo@0.1.3 context frontend
+```
+
+## Outreach Message
+
+```text
+Hey, I’m testing a small local CLI feature for Prismo called PrismoDev.
+
+It scans a repo and local Codex/Claude Code logs to find token waste from context bloat, tool output, oversized instruction files, missing ignore files, and long AI coding sessions.
+
+No API keys, no prompt interception, no subscription proxying.
+
+Could you run these three commands from a real repo and send me a screenshot or notes on what was useful/confusing?
+
+npx getprismo@0.1.3 scan --usage --no-report
+npx getprismo@0.1.3 setup
+npx getprismo@0.1.3 watch --once
+
+I’m mainly trying to learn whether the output is useful enough for developers who use Codex, Claude Code, or Cursor.
+```
+
+## What To Collect
+
+For each tester, record:
+
+- Tool used: Codex, Claude Code, Cursor, other
+- OS: macOS, Linux, Windows/WSL
+- Did all three commands run without help?
+- Did local logs show up?
+- Did `watch --once` show a useful active session?
+- Did they understand exact proxy tracking vs local session estimates?
+- Most surprising/useful output
+- Most confusing output
+- Would they run `scan --fix`?
+- Would they run `optimize`?
+- What did they expect PrismoDev to do next?
+- Did they ask for live reduction, dashboard/team reporting, or exact proxy setup?
+
+## Feedback Form
+
+```text
+Name:
+Repo type:
+Coding tool:
+OS:
+
+Commands completed:
+- scan --usage --no-report: yes/no
+- setup: yes/no
+- watch --once: yes/no
+
+Were local logs found?
+
+Did any output surprise you?
+
+What was confusing?
+
+Do you understand this distinction?
+- Local scan = heuristic repo/context risk
+- Local logs = exact only when logs expose token fields
+- Prismo proxy = exact when traffic uses Prismo base URL
+
+Would you run scan --fix or optimize?
+
+What should PrismoDev do next?
+- better CLI visibility
+- dashboard/team reports
+- exact Codex/Cursor proxy setup
+- live waste reduction
+- other
+
+Screenshot/output notes:
+```
+
+## Success Criteria
+
+Treat the first test loop as successful if:
+
+- 5+ developers run the commands.
+- At least 4 can explain PrismoDev in one sentence.
+- At least 3 find one output useful or surprising.
+- At least 2 ask for one of: live reduction, team dashboarding, exact proxy setup.
+- No more than 1 person gets blocked by install/run friction.
+
+## Decision Rules
+
+After 5-10 user runs, choose the next product bet:
+
+1. If users love CLI visibility: improve `watch` and add local scan history.
+2. If users ask for teams/reporting: update the Prismo Dev dashboard to import new JSON fields.
+3. If users ask for exact tracking: build guided API/base-url setup for Codex/Cursor-compatible tools.
+4. If users ask for reducing waste live: start optional live reduction design.
+
+Default if feedback is mixed:
+
+```bash
+npx getprismo setup codex
+```
+
+This should generate safe instructions for routing API-mode Codex traffic through Prismo. It should not auto-edit user config.
+
+## Scope Guardrails
+
+Do not build these until user validation points there:
+
+- Claude Code hooks
+- MCP interception
+- prompt rewriting
+- shell-output compression
+- dashboard persistence
+- team reporting
+- live reduction
+
+Keep claims honest:
+
+- Exact usage/cost only when traffic flows through Prismo.
+- Local logs are visibility signals, not guaranteed provider billing records.
+- Subscription coding sessions are local-log/heuristic unless the tool supports routed base URL traffic.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "getprismo",
-  "version": "0.1.3",
+  "version": "0.1.4",
   "description": "Local AI coding workflow scanner for Codex, Claude Code, Cursor, and token-waste diagnostics.",
   "license": "UNLICENSED",
   "publishConfig": {
@@ -12,6 +12,7 @@
   "files": [
     "bin/",
     "lib/",
+    "docs/prismodev-user-testing.md",
     "README.md"
   ],
   "engines": {