registry / dsh-tg-bridge / 1.13.1
dsh-tg-bridge@1.13.1
Telegram bridge for DeepSeek Harness: text, voice, photos and project-per-topic sessions without opening the desktop app
AI Security Review
scanned 10d ago · by lpm-firewall-aiSource inspection found no malicious behavior. Scanner-flagged agent configuration, process, and network operations resolve to user-invoked controls and the package’s bridge and maintenance functions.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review; previous stored version diff introduced dangerous source
Trigger
The configured bridge runs through the DSH plugin or an explicit CLI command; persona changes require a /soul message.
Impact
No unrelated credential transfer, unconsented agent configuration change, or other confirmed attack impact was established.
Mechanism
The bridge uses its bot token for Telegram API requests, checks allowed user IDs, and fetches public model catalog data for prices.
AI rationale
Source inspection found no malicious behavior. Scanner-flagged agent configuration, process, and network operations resolve to user-invoked controls and the package’s bridge and maintenance functions.
Network endpoints2
api.telegram.orgmodels.dev
Decision evidence
public snapshotAI called this Clean at 96.0% confidence as Benign with low false-positive risk.
Evidence for AI clean decision
- package.json exposes start, setup, and hooks as explicit commands; it defines no install lifecycle hook.
- bridge.mjs reads the Telegram bot token for Telegram API calls and checks allowed user IDs before handling messages. Its inspected network calls use Telegram; it also fetches public model catalog data in prices.mjs.
- The global agent persona write in control.mjs is reached through the user-entered /soul command, which writes the supplied text to the DSH home instructions file.
- The inspected panel, profile-edit, setup, and restart helpers render chat content, edit profile settings, run setup checks, and restart the bridge; they did not establish an attack surface.
Evidence against
Behavioral surface
ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell
HighEntropyStringsUrlStrings
Source & flagged code
10 flaggedinstall-hooks.mjsView file
13// Usage: node install-hooks.mjs [--uninstall]
L14: import { execFileSync } from 'node:child_process';
L15: import { existsSync, mkdirSync, chmodSync, writeFileSync } from 'node:fs';
High
control.mjsView file
1754try {
L1755: const drives = execFileSync('powershell', ['-NoProfile', '-NonInteractive', '-Command', '(Get-PSDrive C).Free'], {
L1756: encoding: 'utf8',
High
•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation.
ed') }));
}
function sessiontitle(id) {
const record = readjson(join(dsh_home, 'storages', 'session_projcache', 'sessions', `session-${bareid(id)}.json`));
const rows = record?.record?.rows ?? record?.rows ?? {};
const title = rows.title?.val;
if (typeof title === 'string' && title.trim()) return title.trim();
const first = rows.titleinput?.val?.first?.text;
if (typeof first === 'string' && first.trim()) return first.tr
Findings
2 Critical7 High4 Medium4 Low
CriticalHardcoded Runtime Data Exfiltrationbridge.mjs
CriticalPrevious Version Dangerous Deltabridge.mjs
HighChild Processinstall-hooks.mjs
HighShellcontrol.mjs
HighCommand Output Exfiltrationsetup.mjs
HighTrigger Reachable External Ai Agent Control Surface Mutationcontrol.mjs
HighSemantic Analysis Limitedpanel.mjs#virtual:normalized:round1
HighKnown Malware Source Similaritybridge-restart-idle.mjs
HighKnown Malware Source Similarityrestart-after-send.mjs
MediumNetwork
MediumEnvironment Vars
MediumShips Build Helperinstall.sh
Affected versions and remediation
This report applies to dsh-tg-bridge@1.13.1.
See all versions of dsh-tg-bridge for other recorded verdicts.
- Review the evidence and your use of dsh-tg-bridge@1.13.1 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.