The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions in SKILL.md direct the agent to execute shell commands (open on macOS, xdg-open on Linux, and start on Windows) to automatically open the generated report in the system's default browser without separate user confirmation for the shell call.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external job descriptions and resumes to generate match scores and interview questions.
Ingestion points: Job description text or URLs and resume content provided by the user are ingested in SKILL.md and processed via multiple prompt files in the prompts/ directory.
Boundary markers: None identified; instructions do not require the use of delimiters or specific instructions to disregard embedded commands in the untrusted data.
Capability inventory: The skill performs file system writes to the user's desktop and executes shell commands as defined in SKILL.md Step 3.
Sanitization: No input sanitization, escaping, or validation logic is specified in the prompts or skill instructions to handle potentially malicious input strings.
[EXTERNAL_DOWNLOADS]: The agent is instructed in SKILL.md to fetch content from user-provided URLs for job description analysis. Additionally, the HTML report template in examples/offer-strategy-template.html loads font assets from Google Fonts, which is a well-known service.
[DYNAMIC_EXECUTION]: The skill generates an HTML file containing embedded scripts (for Markdown conversion and language toggling) and directs the agent to open this file, resulting in the execution of the generated payload within a browser context.

job-description-skill