The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly mentions the ability to read sensitive files such as .env and access the GEMINI_API_KEY. Instructions that access environment configuration files are high risk as these files frequently store authentication secrets.
[REMOTE_CODE_EXECUTION]: The skill is a wrapper for an external repository (github.com/santifer/career-ops) and provides instructions to run npm install and various Node.js scripts (e.g., scan.mjs, gemini-eval.mjs, update-system.mjs) from this third-party source.
[DATA_EXFILTRATION]: The skill has the combined capability of reading local sensitive credentials (API keys, .env) and making network requests to external recruiting platforms (Greenhouse, Lever, Ashby, Workable, SmartRecruiters, Recruitee) and LLM providers. This combination creates a significant path for data exfiltration.
[COMMAND_EXECUTION]: The instructions include execution of shell commands such as npm, node, and npx playwright. These commands interact directly with the underlying system and can perform a wide range of administrative or operational tasks.
[EXTERNAL_DOWNLOADS]: The skill fetches content from an external GitHub repository and triggers further downloads via package managers (npm) and browser automation tools (playwright install chromium).
[DYNAMIC_EXECUTION]: The skill involves the runtime execution of multiple JavaScript files (.mjs) and the generation of PDFs and reports, which involves compiling or processing data into executable or complex formats at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad attack surface for indirect injection by scanning job portals and recruiting websites, which are attacker-controllable external sources.
Ingestion points: External recruiting portals (Greenhouse, Lever, etc.), cv.md, article-digest.md, and local configuration files.
Boundary markers: None detected; the skill lacks specific delimiters or instructions to ignore embedded commands in the data it processes.
Capability inventory: File writing, shell command execution (Node.js/npm), and network operations to various third-party domains.
Sanitization: No evidence of sanitization, validation, or escaping of the external job portal data before it is processed by the agent.

career-ops