The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the user or agent to execute commands that modify local system configurations, such as editing ~/.cursor/mcp.json and running agent-specific setup commands like claude mcp add and code --add-mcp.
[REMOTE_CODE_EXECUTION]: The setup instructions include the use of npx -y to download and execute packages such as add-mcp and skills (from the techjobs-vn/skills repository) to automate configuration and install additional functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, including candidate messages and CV text, when referrers review requests. This content could potentially contain instructions aimed at manipulating the agent's behavior.
Ingestion points: External data is ingested via list_referrer_requests, get_referrer_request_cv, and match_cv as described in SKILL.md.
Boundary markers: The skill includes explicit defensive instructions in the 'Rules' section, advising the agent to treat these inputs as data only and to ignore any embedded instructions.
Capability inventory: The skill possesses the ability to modify user CVs (update_cv_variant), send referral messages (request_referral), and update application tracking statuses (track_job).
Sanitization: The mitigation relies on the agent's adherence to natural language boundary instructions rather than programmatic sanitization of the incoming text.

techjobs