The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains a hardcoded authentication token (RUBE_TOKEN) in scripts/create_sheet.py, scripts/create_sheet_mcp.py, and scripts/search_jobs.py. This token is a valid JSON Web Token (JWT) that grants access to the Rube/Composio platform and should be managed via environment variables rather than hardcoded in the source code.
[DYNAMIC_EXECUTION]: The scripts create_sheet_mcp.py and search_jobs.py dynamically assemble Python code blocks as strings and transmit them to a remote server for execution via the RUBE_REMOTE_WORKBENCH tool. This pattern of remote code execution (RCE) via a workbench tool is a high-capability feature that poses a security risk if the generated code is influenced by untrusted inputs.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection:
Ingestion points: The main script (scripts/search_jobs.py) fetches job titles, descriptions, and company metadata from LinkedIn via the Apify API.
Boundary markers: There are no delimiters or explicit instructions provided to the agent to treat the ingested job description content as untrusted data.
Capability inventory: The skill can execute remote code via Rube, perform authenticated API requests to Apify, and modify Google Sheets.
Sanitization: The skill performs no sanitization or escaping of the job description text, which is directly summarized and presented to the agent, potentially allowing an attacker to embed malicious instructions in a job posting.

job-posting-intent