The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The installation instructions require cloning a repository from an untrusted source (https://github.com/santifer/career-ops.git).
[REMOTE_CODE_EXECUTION]: The skill instructs the user to execute downloaded code through npm install, go build, and by running a bash script (batch-runner.sh) that orchestrates sub-agents.
[COMMAND_EXECUTION]: The skill relies on significant local shell operations, including compiling a Go binary, running Node.js scripts, and executing parallel sub-agents using the claude CLI.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process job descriptions from external, untrusted sources (e.g., job portal URLs). This creates a surface for malicious instructions hidden in job listings to influence the agent's behavior.
Ingestion points: External job URLs and raw job description text pasted into the session.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands in the job data are documented in the skill structure.
Capability inventory: The skill writes to the file system (TSV logs, Markdown reports, PDFs) and executes shell commands via the Claude CLI.
Sanitization: There is no evidence of sanitization or filtering of the job description text before it is processed by the AI.
[DYNAMIC_EXECUTION]: The skill uses Playwright to dynamically render HTML templates into PDF files, which involves a complex browser-based execution environment that processes injected keywords from external job descriptions.

career-ops-job-search