📦
Audit History
linkedin-easy-apply - 4 audits
Version comparison
Capability and finding changes across audited versions, newest first.
Audit version 4 Latest
Jul 5, 2026, 08:07 PM
Most shell-backtick detections are false positives caused by Markdown fences and inline code. Confirmed risks include sandbox-disabled Chromium, persistent browser session storage, /tmp state logs, and automated personal data submission. No prompt injection or credential exfiltration instructions were found.
1
Files scanned
175
Lines analyzed
14
Review items
0
False positives ignored
Confirmed security concerns (3)
High
Environment file access
userDataDir: process.env.CHROME_PROFILE || `${process.env.HOME}/.cache/linkedin-chrome`,
The line configures a persistent Chrome user data directory that can hold authenticated LinkedIn session data. Reusing this profile is sensitive even without direct exfiltration.
High
UID/GID manipulation
args: ['--no-sandbox', '--disable-setuid-sandbox', '--disable-dev-shm-usage', '--start-maximized']
The Chromium launch arguments disable the browser sandbox and setuid sandbox. This weakens process isolation if LinkedIn or any loaded content is compromised.
Medium
Automated Personal Data Submission
The skill directs an agent to upload a resume and submit LinkedIn applications. This can expose personal data or submit unwanted applications if approval controls are weak.
The workflow explicitly searches LinkedIn Easy Apply, uploads a verified resume, and clicks submit when required fields are known. The context confirms automated third-party submission of personal job application data.
Capability review items (7)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
Medium
Hidden file access
CHROME_PROFILE=$HOME/.cache/linkedin-chrome
The skill recommends a persistent hidden Chrome profile under the user home directory. That profile can retain LinkedIn session cookies and other browser state.
Medium
Hidden file access
userDataDir: process.env.CHROME_PROFILE || `${process.env.HOME}/.cache/linkedin-chrome`,
Puppeteer is configured to reuse a hidden browser profile path. Persistent profiles can expose authenticated session data if permissions or isolation are weak.
Medium
Temp directory access
STATE_DIR=/tmp/linkedin-easyapply-daily
The suggested state directory is under /tmp. Application state in a shared temporary path can be read or tampered with on multi-user systems.
Medium
Temp directory access
/tmp/linkedin-easyapply-daily/state.json
The state file path is under /tmp and can contain application status. Shared temporary storage creates privacy and integrity risk.
Medium
Temp directory access
/tmp/linkedin-easyapply-daily/results.jsonl
The results log path is under /tmp and can contain job URLs and submission outcomes. Shared temporary storage creates privacy and tampering risk.
Medium
Temp directory access
State: /tmp/linkedin-easyapply-daily/state.json
The reporting section exposes the state path under /tmp. That location is a weak default for user-specific application history.
Medium
Temp directory access
Log: /tmp/linkedin-easyapply-daily/results.jsonl
The reporting section exposes the log path under /tmp. Logs may include personal job search activity and should not use shared temporary storage.
Risk Factors
⚙️ External commands (19)
🌐 Network access (1)
📁 Filesystem access (7)
🔑 Env variables (4)
Detected Patterns
UID/GID manipulation
Audited by: codex
Jul 5, 2026, 08:07 PM
Most shell-backtick detections are false positives caused by Markdown fences and inline code. Confirmed risks include sandbox-disabled Chromium, persistent browser session storage, /tmp state logs, and automated personal data submission. No prompt injection or credential exfiltration instructions were found.
1
Files scanned
175
Lines analyzed
14
Review items
0
False positives ignored
Confirmed security concerns (3)
High
Environment file access
userDataDir: process.env.CHROME_PROFILE || `${process.env.HOME}/.cache/linkedin-chrome`,
The line configures a persistent Chrome user data directory that can hold authenticated LinkedIn session data. Reusing this profile is sensitive even without direct exfiltration.
High
UID/GID manipulation
args: ['--no-sandbox', '--disable-setuid-sandbox', '--disable-dev-shm-usage', '--start-maximized']
The Chromium launch arguments disable the browser sandbox and setuid sandbox. This weakens process isolation if LinkedIn or any loaded content is compromised.
Medium
Automated Personal Data Submission
The skill directs an agent to upload a resume and submit LinkedIn applications. This can expose personal data or submit unwanted applications if approval controls are weak.
The workflow explicitly searches LinkedIn Easy Apply, uploads a verified resume, and clicks submit when required fields are known. The context confirms automated third-party submission of personal job application data.
Capability review items (7)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
Medium
Hidden file access
CHROME_PROFILE=$HOME/.cache/linkedin-chrome
The skill recommends a persistent hidden Chrome profile under the user home directory. That profile can retain LinkedIn session cookies and other browser state.
Medium
Hidden file access
userDataDir: process.env.CHROME_PROFILE || `${process.env.HOME}/.cache/linkedin-chrome`,
Puppeteer is configured to reuse a hidden browser profile path. Persistent profiles can expose authenticated session data if permissions or isolation are weak.
Medium
Temp directory access
STATE_DIR=/tmp/linkedin-easyapply-daily
The suggested state directory is under /tmp. Application state in a shared temporary path can be read or tampered with on multi-user systems.
Medium
Temp directory access
/tmp/linkedin-easyapply-daily/state.json
The state file path is under /tmp and can contain application status. Shared temporary storage creates privacy and integrity risk.
Medium
Temp directory access
/tmp/linkedin-easyapply-daily/results.jsonl
The results log path is under /tmp and can contain job URLs and submission outcomes. Shared temporary storage creates privacy and tampering risk.
Medium
Temp directory access
State: /tmp/linkedin-easyapply-daily/state.json
The reporting section exposes the state path under /tmp. That location is a weak default for user-specific application history.
Medium
Temp directory access
Log: /tmp/linkedin-easyapply-daily/results.jsonl
The reporting section exposes the log path under /tmp. Logs may include personal job search activity and should not use shared temporary storage.
Risk Factors
⚙️ External commands (19)
🌐 Network access (1)
📁 Filesystem access (7)
🔑 Env variables (4)
Detected Patterns
UID/GID manipulation
Audited by: codex
Jun 30, 2026, 10:56 AM
Static code-execution and weak-crypto alerts are mostly false positives caused by Markdown backticks and keyword matching. The skill still has medium risk because it guides authenticated LinkedIn browser automation, persistent profile use, network access, and sandbox-disabled Chromium execution.
1
Files scanned
175
Lines analyzed
9
Review items
1
False positives ignored
Capability review items (5)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
Medium
Authenticated Browser Automation With Account Actions
The skill directs an agent to use a logged-in LinkedIn browser session and submit applications. This is user-authorized automation, but it can act on an account and should require explicit operator control.
The instructions explicitly launch a browser profile and click through application submission steps. No evidence found of credential theft or unauthorized exfiltration.
Medium
Persistent Browser Profile May Contain Session Data
The skill recommends a persistent Chrome profile for manual LinkedIn login. That profile can contain cookies and session data, so storage permissions and cleanup matter.
The profile path is explicit and tied to authenticated browser use. The file does not instruct copying or transmitting the profile.
Medium
Chromium Sandbox Disabled
The Puppeteer launch arguments disable Chromium sandbox protections. This can increase impact if a browser exploit or malicious page is encountered.
The launch arguments directly include sandbox-disabling flags. This is a common container workaround, but it weakens browser isolation.
Low
Expected LinkedIn Network Access
The hardcoded URL points to LinkedIn job application pages, which matches the stated purpose. No evidence found of third-party exfiltration endpoints.
The only observed hardcoded network destination is LinkedIn. This is expected for an Easy Apply automation skill.
Low
Temporary Logs May Contain Application History
The skill stores state and result logs under /tmp. These logs can expose job targets and application outcomes on shared systems.
The log and state paths are explicit. The exact logged fields are not implemented in this file, so the data exposure risk is limited.
Static false positives ignored (1)
These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.
Low
Static Analyzer Keyword Hits Are False Positives
The Ruby backtick, weak-crypto, sensitive environment-file, and system-reconnaissance alerts come from Markdown formatting or ordinary text. No evidence found of those behaviors.
The referenced lines are Markdown examples, process environment configuration, or plain prose. They do not define Ruby execution, cryptography, .env access, or system reconnaissance.
Risk Factors
⚙️ External commands (1)
🌐 Network access (1)
📁 Filesystem access (3)
🔑 Env variables (1)
Detected Patterns
Sandbox-Disabled Browser Launch PatternPersistent Authenticated Profile PatternAutomated Submission Pattern
Audited by: codex
Apr 30, 2026, 12:27 PM
Skill is a documentation/guide template for building LinkedIn Easy Apply automation. Static scanner flagged 48 potential issues but most are false positives from code examples in documentation. Confirmed legitimate browser automation patterns. Primary concerns are (1) LinkedIn Terms of Service violations for automated job applications, (2) pattern detection for Puppeteer/Chrome flags including --disable-setuid-sandbox which is a standard Chrome launch option. Skill explicitly avoids credential storage and includes conservative answer guardrails.
1
Files scanned
175
Lines analyzed
9
Review items
0
False positives ignored
Capability review items (4)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
LinkedIn Terms of Service Violation Risk
Automated job application submissions likely violate LinkedIn's Terms of Service. Users risk account suspension or legal action from LinkedIn. This is a business/legal risk rather than technical security risk.
Skill is explicitly designed to automate LinkedIn Easy Apply, which LinkedIn prohibits. Line 42 mentions stopping for CAPTCHA but does not address ToS compliance.
Medium
Chrome Security Flags Misidentified
Static scanner flagged --disable-setuid-sandbox as UID/GID manipulation. This is a standard Chrome launch flag for containerized environments, not malicious UID manipulation.
The --disable-setuid-sandbox flag is documented Chromium behavior for running in Docker containers. No actual UID/GID changes occur.
Medium
Code Examples Flagged as External Commands
Static scanner detected backtick execution patterns in code examples. These are documentation examples, not actual executions in the skill.
All external command detections are in code blocks showing example Puppeteer/JavaScript patterns. Skill is a documentation template, not executable code.
Low
Environment Variable Access for Configuration
Skill uses environment variables for configuration (RESUME_PDF, CHROME_PROFILE, STATE_DIR). This is standard configuration pattern, not credential exposure.
Environment variables are documented configuration values, not secrets. Line 25 explicitly states skill contains no credentials.
Risk Factors
⚙️ External commands (21)
SKILL.md:49-59 SKILL.md:59-63 SKILL.md:63-69 SKILL.md:69-82 SKILL.md:82-88 SKILL.md:88-90 SKILL.md:90-99 SKILL.md:99-100 SKILL.md:100-102 SKILL.md:102-104 SKILL.md:104-132 SKILL.md:132 SKILL.md:132 SKILL.md:132-137 SKILL.md:137-144 SKILL.md:144-148 SKILL.md:148-151 SKILL.md:151-157 SKILL.md:157-166 SKILL.md:166-172 SKILL.md:172
🌐 Network access (1)
📁 Filesystem access (7)
🔑 Env variables (4)
⚡ Contains scripts (1)
Audited by: claude