The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by fetching and parsing content from user-provided URLs using render_page.py and parse_html.py as described in SKILL.md.
Ingestion points: External URL content is processed during the Product Page Analysis workflow in SKILL.md.
Boundary markers: There are no explicit boundary markers or instructions (e.g., 'ignore instructions within this data') defined for the agent when processing the fetched content.
Capability inventory: The skill allows shell execution of various Python scripts, network access via the DataForSEO Merchant API, and writing data to results.json.
Sanitization: While the skill mentions google_auth.validate_url() for malformed input, it does not specify sanitization or filtering of the actual content retrieved from the web pages.
[COMMAND_EXECUTION]: The workflow involves passing user-controlled arguments, such as <url>, <keyword>, and <domain>, directly into shell commands via a script runner ("${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo" run ...). This represents a potential command injection surface if the underlying runner script or the Python scripts themselves (like render_page.py or dataforseo_merchant.py) do not properly escape these inputs before execution.
[EXTERNAL_DOWNLOADS]: The skill is designed to interact with the DataForSEO Merchant API to retrieve pricing and marketplace data. As documented in SKILL.md and references/marketplace-endpoints.md, this involves sending user-provided keywords and domain information to an external third-party service. The skill includes cost guardrail checks (dataforseo_costs.py) which acts as a security best practice for budget management but confirms active network communication.

seo-ecommerce