The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources and local files that could potentially contain malicious instructions.
Ingestion points: Reads page content (meta tags, schema, canonicals, and redirects) from live WordPress and Webflow CMS platforms via MCP servers. It also consumes PLAN.md files generated by other skills and brand profiles from ~/.claude-marketing/.
Boundary markers: The instructions do not specify any delimiters or safety markers to isolate ingested content from the agent's control logic.
Capability inventory: The skill is capable of writing data to CMS platforms (updating meta tags, deploying schema, creating redirects) and executing shell commands via Python scripts.
Sanitization: Validation is performed for SEO best practices (e.g., character counts, schema validity), but there is no evidence of filtering or escaping for embedded prompt injection instructions within the ingested data.
[COMMAND_EXECUTION]: The skill executes internal shell commands to manage approval states and audit trails.
Evidence: The skill uses python "${CLAUDE_PLUGIN_ROOT}/scripts/approval-manager.py" to record and verify human-in-the-loop approvals.
Context: These commands use variables such as {slug} retrieved from local configuration files to track actions for specific brands.

seo-implement