The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill audits external web pages, creating a surface for indirect prompt injection.
Ingestion points: Audits data from rendered_dom and raw_html in SKILL.md.
Boundary markers: Absent; no delimiters or instructions to ignore embedded instructions in audited pages are specified.
Capability inventory: The skill has access to high-privilege tools including Bash, WebFetch, Read, Grep, and Glob.
Sanitization: No sanitization or escaping procedures are defined for data extracted from external pages before it is passed to other tools or included in generated JSON-LD.
[COMMAND_EXECUTION]: The skill executes shell commands via the Bash tool using parameters derived from external page content.
Evidence: node ${CLAUDE_SKILL_DIR}/../../scripts/validate-jsonld.mjs --url <u> in SKILL.md.
The <u> placeholder represents a URL extracted from the audited page, which could lead to command injection if the agent fails to sanitize the input before shell execution.
[DYNAMIC_EXECUTION]: The skill uses dynamic paths involving environment variables and directory traversal to locate and execute local scripts.
Evidence: Execution of ${CLAUDE_SKILL_DIR}/../../scripts/validate-jsonld.mjs in SKILL.md.

seo-local