The Agent Skills Directory

External Network Requests: The skill performs network operations using curl to fetch site metadata and sample page content. These requests are directed to domains specified by the user or discovered during the audit process, which is the primary purpose of the skill.
Local Command Execution: The skill utilizes a bundled shell script to perform technical probes. This script leverages common system utilities to analyze the structure and delivery of the targeted documentation without requiring elevated privileges.
Indirect Prompt Injection Surface: The skill processes external HTML and Markdown content to evaluate its semantic structure. This creates a potential surface where instructions embedded in the audited documentation could attempt to influence the agent's behavior.
Ingestion points: Untrusted website content is fetched in the baseline probes script and returned to the agent context for evaluation.
Boundary markers: The instructions do not define explicit delimiters to separate the site content from the agent's auditing logic.
Capability inventory: The skill can perform network requests and execute local shell commands via its bundled scripts.
Sanitization: The script extracts structural elements, but the content is evaluated by the agent without specific sanitization for natural language instructions.

agent-readiness-audit