Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local Upwork proposal assistant that writes application and hook-history files, with no evidence of exfiltration, privilege escalation, hidden execution, or destructive behavior.
Install only if you are comfortable storing freelancer profile details, generated proposals, job history, and winning hooks in local workspace files. Review generated proposals before any real Upwork submission, and treat the skill's auto-apply language as local drafting/logging unless you separately connect it to a real Upwork workflow.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
# New entries appended by HunterAI after each run.
# Template entry structure below — do not delete this block.
<!--
═══════════════════════════════════════════════════
JOB ID: UPW-YYYYMMDD-XXXX
TITLE: [Job Title Here]
The README markets automated Upwork job application behavior and describes persistent logging to local files, but it does not clearly warn users about the operational, policy, and privacy risks of automated proposal submission or long-term storage of proposal/job-history data. In this context, the omission is security-relevant because users may trigger account-impacting automation and create sensitive local records without informed consent or safeguards.
The skill gives a specific example trigger, but then broadens it with "or similar," without defining acceptable variants or exclusions. That creates an unclear activation boundary and increases the chance of unintended invocation from loosely related user requests.
The skill instructs the agent to append full application data, including proposal text, to a persistent log immediately and without requiring explicit user confirmation or a prior warning about file modification. This creates a silent persistence channel that can store sensitive business content or user-authored material unexpectedly, increasing privacy, integrity, and workspace-tampering risk.
This flow persists derived content into assets/PROPOSAL_VAULT.md and updates APPLICATION_LOG.md without a clear warning that durable project files will be changed. Because the vault is treated as future prompt material, silent writes can poison later outputs, preserve sensitive client/job details, and alter agent behavior across sessions.
The status update path changes persistent records and can automatically trigger additional writes through Loop B, but the user is not clearly informed that one command may cascade into multiple file modifications. This increases the risk of unintended state changes, hidden persistence, and integrity issues in the workspace, especially because the auto-trigger expands the original action scope.
No suspicious patterns detected.