registry / @vibes.diy/api-svc / 14.3.141
@vibes.diy/api-svc@14.3.141
⚠ Under reviewWebSocket-based API for vibes.diy chat persistence and app deployment. Supports both HTTP POST and WebSocket connections using the same handler via the Evento event-driven framework.
AI Security Review
scanned 18d ago · by lpm-firewall-aiThe inspected flagged source shows an explicit, write-gated avatar migration, not a confirmed attack. I cannot mark the package clean because the required clean citation coverage exceeds the 128-reference limit, so manual review is appropriate.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
An operator runs the admin backfill command; database and storage writes require --apply.
Impact
The inspected behavior updates avatar assets and settings as part of the migration; no unrelated credential or application-record transfer was established.
Mechanism
The script fetches existing avatar images, transforms them, and stores them in the configured R2 bucket.
AI rationale
The inspected flagged source shows an explicit, write-gated avatar migration, not a confirmed attack. I cannot mark the package clean because the required clean citation coverage exceeds the 128-reference limit, so manual review is appropriate.
Decision evidence
public snapshotAI called this Manual Review at 78.0% confidence as Unknown with medium false-positive risk.
AI review observations
- The manifest identifies index.js as the package entry point.
- The avatar backfill is exposed as an explicit admin command.
- The backfill fetches stored avatar assets from the configured asset host.
- Its write path stores transformed avatar bytes in the R2 bucket.
- The --apply flag selects the admin database connection; without it, the script reports that nothing is written.
Evidence against
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetwork
HighEntropyStringsUrlStrings
Source & flagged code
5 flaggedusage-report/backfill-avatars.jsView file
1import { spawnSync } from "node:child_process";
L2: import * as fs from "node:fs";
...
L24: NEON_DATABASE_ADMIN_URL write connection (required for --apply)
L25: ASSET_FETCH_BASE public host serving /assets/cid/ (default https://vibes.diy)
L26: --apply additionally shells out to \`wrangler r2 object put\` — wrangler must be
...
L37: throw new Error(`fetch ${row.getURL} returned ${res.status}`);
L38: const buf = Buffer.from(await res.arrayBuffer());
L39: if (buf.byteLength !== row.size) {
...
L92: return Result.Err(`wrangler r2 object get exited ${res.status}`);
L93: if (res.stdout.equals(work.newBytes) === false)
L94: return Result.Err("stored avatar differs from transformed bytes");
...
L140: const apply = args.includes("--apply");
Critical
Hardcoded Runtime Data Exfiltration
Source sends credentials or rich application records to a package-controlled external receiver enabled by default.
usage-report/backfill-avatars.jsView on unpkg · L1usage-report/r2-affected-apps.shView file
•path = usage-report/r2-affected-apps.sh
kind = build_helper
sizeBytes = 2493
magicHex = [redacted]
Medium
Ships Build Helper
Findings
1 Critical2 High4 Medium5 Low
CriticalHardcoded Runtime Data Exfiltrationusage-report/backfill-avatars.js
HighSemantic Analysis Limitedintern/design-repair.js#virtual:normalized:round1
HighPrevious Version Dangerous Deltaintern/opens-check.js
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperusage-report/r2-affected-apps.sh
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
LowAi Review Evidencepackage.json
Affected versions and remediation
This report applies to @vibes.diy/api-svc@14.3.141.
See all versions of @vibes.diy/api-svc for other recorded verdicts.
- Review the evidence and your use of @vibes.diy/api-svc@14.3.141 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.