registry / skalpel / 4.0.54

skalpel@4.0.54

⚠ Under review

Behavioral graph for AI-assisted coding — learns how you work and steers Claude Code + Codex in real time.

Static Scan Results

scanned 1mo ago · by rust-scanner

Static analysis flagged 16 finding(s) at 86.0% confidence. This version is warn-only unless an AI or security-team review confirms malicious behavior.

Static reason

High-risk behavior combination matched malicious policy.

Decision evidence

public snapshot

Behavioral surface

Source

ChildProcessCryptoEnvironmentVarsEvalFilesystemNetworkShell

Supply chain

HighEntropyStringsUrlStrings

Manifest

NoLicense

scanned 34 file(s), 765 KB of source, external domains: accounts.google.com, app.example.com, cli.skalpel.ai, err.example.com, gone.example.com, graph.skalpel.ai, nope.example.com, ok.example.com, secure.example.com, skalpel-prod.auth.us-west-2.amazoncognito.com, x.example.com

Source & flagged code

7 flagged

package.jsonView file

Lines 1-29json

2 "name": "skalpel",

3 "version": "4.0.54",

4 "description": "Behavioral graph for AI-assisted coding — learns how you work and steers Claude Code + Codex in real time.",

5 "type": "module",

6 "bin": {

7 "skalpel": "skalpel-setup.mjs"

8 },

9 "scripts": {

10 "postinstall": "node postinstall.mjs || true"

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L10

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg · L10

11 },

12 "files": [

13 "*.mjs",

14 "README.md"

15 ],

16 "engines": {

17 "node": ">=18"

18 },

19 "keywords": [

20 "claude-code",

21 "codex",

22 "ai",

23 "productivity",

24 "behavioral-analytics"

25 ],

26 "license": "UNLICENSED",

27 "homepage": "https://skalpel.ai"

28}

anchor-shadow.mjsView file

Lines 1-39javascript

1#!/usr/bin/env node

2// anchor-shadow.mjs — ship-status "Auto-Run-the-Anchor", SHADOW mode, READ-ONLY.

3//

4// WHAT: when the user falls into the cycle-12 signature (>=2 consecutive terse re-verify/doubt turns

5// about the SAME claim — "did it merge?", "you sure it's live?", "did that actually push?") AND that

6// doubt is about SHIP STATUS (merged / pushed / live / deployed / price / trial / CI / passing /

7// released), reconstruct the concrete anchor from recent session + repo context (a PR number, a URL,

8// a commit sha, a Stripe price), run ONE read-only allowlisted probe to resolve the doubt, and LOG

9// what we WOULD have injected. It NEVER injects into the live model and NEVER mutates anything.

Lines 1-22javascript

1#!/usr/bin/env node

2// skalpel login — Sign in with Google, fully through AWS (no localhost).

3// this CLI → browser → Google → https://graph.skalpel.ai/oauth/callback (AWS) → we poll AWS for

4// the token → save it. The server holds the client secret and does the code exchange; we only ever

5// see the finished Google id_token. Branded onboarding reproduces the original prosumer `setup.mjs`

6// red block wordmark 1:1. Run: `node skalpel-login.mjs`.

7import { readFileSync, writeFileSync, mkdirSync, realpathSync } from "node:fs";

8import { homedir }

efficacy-eval.mjsView file

Lines 220-260javascript

220 .trim();

221 return "";

222}

223

224// The normalized proof-command STRING, derived identically to how the confirmed-lie record derives it

225// (`[cmd, ...args].join(" ")`), so a later PASS of the same command matches by exact string in

226// buildFixDeltas. Returns null if the entry ran no allowlisted proof.

227function proofCommandString(

Source & flagged code

7 flagged · loading source

package.jsonView file

Static Scan Results

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code