The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from external job postings fetched via browser tools or user input. This creates an Indirect Prompt Injection surface where malicious instructions in a job description could attempt to influence the agent's behavior.
Ingestion points: Job posting content retrieved in "SKILL.md" (Step 1) and "references/browser-setup.md" using browser automation tools.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the fetched job data.
Capability inventory: File read/write access to the ".proficiently/" directory and browser navigation/extraction tools.
Sanitization: No input validation or sanitization is performed on the job posting text before it is analyzed by the writing agent.
[COMMAND_EXECUTION]: Employs browser automation tools (e.g., "navigate", "javascript_tool", "get_page_text") as described in "references/browser-setup.md" to visit and extract data from external websites. While these are used for the skill's primary function, they provide the agent with interaction capabilities with the browser environment.
[DATA_EXFILTRATION]: The skill accesses a hidden directory in the user's home folder ("~/.proficiently/") to store and retrieve personal data such as resumes and work history profiles. Access to hidden folders in the home directory is sensitive, although here it appears to be the designated storage path for the application's data ecosystem.

cover-letter