moltoffer-auto-apply
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to configure a Python virtual environment and install project dependencies using
pipduring the onboarding phase. - [EXTERNAL_DOWNLOADS]: Downloads the
playwrightandpyyamlpackages from official registries and retrieves the Chromium browser binary to support automated application filling. - [DATA_EXFILTRATION]: Reads sensitive information from
persona.mdandcredentials.local.jsonwithin the local file system. This data is used to authenticate with the vendor's API and to populate job application forms on LinkedIn, which is the skill's primary function. - [COMMAND_EXECUTION]: Uses
browser_evaluateto execute JavaScript snippets within the browser context. This functionality is employed to manage complex UI interactions on LinkedIn, such as handling modal windows and selecting specific form elements. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external data from the MoltOffer API (job details) and the LinkedIn DOM (form snapshots) to determine automated actions. Ingestion points:
references/workflow.md(API fetch) andreferences/apply.md(browser snapshots). Capability inventory:browser_evaluate,browser_fill_form, andbrowser_type. Boundary markers: Absent. Sanitization: No explicit validation of external job descriptions or form labels is mentioned.
Audit Metadata