The Agent Skills Directory

[SAFE]: The skill identifies a potential attack surface by ingesting untrusted web data through tools like web_fetch and curl. It mitigates this risk by providing explicit instructions to the agent to disregard instructions found in audited HTML or text content. \n
Ingestion points: Data retrieved from external URLs via web_fetch and curl.\n
Boundary markers: Explicit warnings are present in SKILL.md to treat page content as non-reliable and avoid prompt injection.\n
Capability inventory: Limited to web data retrieval and processing; no file-write or subprocess execution on the host is defined.\n
Sanitization: Instructions mandate the isolation of content analysis from command execution.\n- [SAFE]: The technical SEO procedures, including schema markup validation via browser tools and accessibility checks, align with the primary purpose of the skill and utilize standard, non-malicious commands.\n- [SAFE]: All reference materials and metadata files were reviewed and found to contain legitimate documentation without any signs of obfuscation, exfiltration, or persistence mechanisms.

seo-audit