@atedone/workspace-context@0.0.22
Workspace shell session refresh integration helper for JavaScript tools.
OpenSSF/OSV Malicious Advisory
imported 2d ago · by OpenSSF/OSVOpenSSF/OSV advisory MAL-2026-18517 confirms this npm version as malicious.
Advisory
MAL-2026-18517
Source
OpenSSF Malicious Packages via OSV
Summary
Malicious code in @atedone/workspace-context (npm)
Details
On require(), the package's top-level state.initialize() writes shell integration files into ~/.config/workspace-context (init.zsh,.zshrc, session-runtime.js, presentation.js, channel.json) and spawns a detached, unref'd Node worker that launches `/bin/zsh -i` with ZDOTDIR pointed at that directory. The injected init.zsh installs a zsh precmd hook that re-executes a sibling script on every prompt, giving a persistent shell foothold triggered by merely importing the module. session.js fetches http://150.136.8.159:8080/workspace/profile over plaintext HTTP from a hardcoded bare IP, appending?username=<USER> read from env.USER/LOGNAME, writes the response to session-profile.json, and presentation.js injects fields from that remote JSON directly into the user's live zsh PROMPT via %F{tone}text%f substitution — remote attacker-controlled content is rendered through zsh prompt expansion on every command. Sensitive API names are reassembled at runtime from string fragments to defeat static analysis: ['bin','ding'].join(''), ['sp','awn','_','sync'].join(''), ['sp','awn'].join(''), ['child','_process'].join(''), ['fe','tch'].join(''), and the env-var name ['WORKSPACE','_CONTEXT','_CHANNEL'].join(''). Access to child_process/fetch is funneled through Reflect.get/Reflect.apply with no plain literal references anywhere. refreshShellIntegration obtains the internal process.binding('spawn_sync') handle to execute `node lib/sync.js`, bypassing the public child_process module. The package has no legitimate need, consistent with its stated purpose, to spawn an interactive shell at import time, hook zsh precmd, or fetch remote content from a bare IP over cleartext HTTP.
Independent LPM AI assessment
SuspiciousIndependent AI assessment: suspicious. LPM AI assessed this version as suspicious; the assessments disagree.
The profile channel is a documented feature, but profile text is inserted into a zsh prompt without escaping and the integration enables prompt_subst. A malicious or compromised configured channel can therefore execute shell commands when the prompt refreshes, so this is a serious vulnerability rather than confirmed malicious behavior.
Static reason
No blocking static signals were detected.
Trigger
Import initializes the integration; when a channel is configured, the shell prompt refresh fetches the profile and renders it.
Impact
A malicious or compromised profile source can execute commands as the user running the shell.
Mechanism
The renderer inserts profile-provided text into its prompt output, while the generated hook enables zsh prompt_subst, which evaluates command substitutions in that text.
AI rationale
The profile channel is a documented feature, but profile text is inserted into a zsh prompt without escaping and the integration enables prompt_subst. A malicious or compromised configured channel can therefore execute shell commands when the prompt refreshes, so this is a serious vulnerability rather than confirmed malicious behavior.
Evidence
index.jslib/session.jslib/presentation.jslib/sync.jsREADME.md.config/workspace-context/channel.json.config/workspace-context/session-profile.json.config/workspace-context/init.zsh.workspace-context/state.json
Independent scan evidence
public snapshotAI called this Suspicious at 96.0% confidence as Critical Vulnerability with medium false-positive risk.
Evidence for AI warning decision
- Importing the package immediately calls initialize().
- The session runtime fetches the configured profile and saves its JSON locally.
- Profile segment text is interpolated directly into the generated prompt.
- The generated shell hook enables prompt_subst before assigning the rendered prompt, allowing command substitutions in profile text to execute in the shell.
- The README says users configure a profile channel before import or call refresh with a channel source.
Evidence against
- The package is described as a shell session and profile channel integration helper.
- The vulnerable prompt path requires a configured profile channel and runs when the shell refresh hook executes.
Behavioral surface
Filesystem
HighEntropyStrings
Source & flagged code
7 flaggedlib/sync.js#virtual:normalized:round1View file
•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 1
High
Semantic Analysis Limited
A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.
lib/sync.js#virtual:normalized:round1View on unpkgindex.jsView file
1'use strict';
const state = require('./lib/state');
const initialized = state.initialize();
module.exports = {
initialized,
refresh: state.initialize,
channelFileName
High
lib/session.jsView file
Findings
7 High2 Low
HighSemantic Analysis Limitedlib/sync.js#virtual:normalized:round1
HighAi Review Evidenceindex.js
HighAi Review Evidencelib/session.js
HighAi Review Evidencelib/session.js
HighAi Review Evidencelib/presentation.js
HighAi Review Evidencelib/sync.js
HighAi Review EvidenceREADME.md
LowFilesystem
LowHigh Entropy Strings
Affected versions and remediation
This report applies to @atedone/workspace-context@0.0.22.
See all versions of @atedone/workspace-context for other recorded verdicts.
- Avoid installing @atedone/workspace-context@0.0.22. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Source advisory published: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.