The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like curl, grep, and tee to interact with external websites and process response data. While these are necessary for the skill's primary purpose of header analysis, they represent a functional capability that acts on user-provided inputs.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches and processes HTTP headers from an arbitrary external TARGET_URL without sanitization or explicit instruction boundaries.
Ingestion points: HTTP response headers are fetched via curl and saved to /tmp/headers_raw.txt (SKILL.md).
Boundary markers: None are present; the agent processes the raw content of the headers.
Capability inventory: The skill uses curl, grep, and cat to retrieve and display content (SKILL.md).
Sanitization: There is no sanitization or filtering of the header values before they are presented to the agent, allowing a malicious server to potentially include instructions within headers like X-Comments or Server that could influence subsequent agent reasoning.

security-headers