The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses the @nomadamas/k-skill package from the npm registry to execute its primary logic and fetch instructions. This allows the execution of code not directly contained within the skill's source files.
[EXTERNAL_DOWNLOADS]: Core components, including the instruction set and the matching execution environment, are downloaded at runtime using the npx command.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted resume data through text inputs or file paths, which could potentially influence agent behavior if malicious instructions are embedded within the resume content.
Ingestion points: Resume data enters the system via the resume_text and resume_file arguments processed in instruction.md and scripts/job_posting_match.py.
Boundary markers: The instructions do not specify explicit delimiters or "ignore" warnings when passing user-provided resume content to the execution script.
Capability inventory: The skill is granted network access to query external job boards (JobKorea and Saramin) and read access to the local file system.
Sanitization: The implementation in scripts/job_posting_match.py includes a clean_text function that sanitizes input by removing HTML, script, and style tags.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands, specifically using npx, to perform its core job matching and information retrieval tasks.

job-posting-match