The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute specific Python scripts located within the plugin's root directory to interface with search APIs.
Evidence: Templates in SKILL.md such as python3 "${CLAUDE_PLUGIN_ROOT}/scripts/connectors/firecrawl.py" search "<keyword>".
Context: The command execution is localized to the plugin's own connector scripts and follows standard patterns for agent-tool interactions.
[PROMPT_INJECTION]: The skill ingests untrusted data from the web (via the WebFetch tool), which presents an indirect prompt injection surface.
Ingestion points: The content of top-ranking URLs fetched during analysis as described in SKILL.md.
Boundary markers: The skill includes a 'Security boundary' section in SKILL.md that explicitly instructs the agent to treat fetched content as evidence only and to ignore any prompt-like directives or owner overrides within that data.
Capability inventory: The agent can execute local connector scripts and write analysis data to memory/research/.
Sanitization: Relies on explicit instruction-following boundaries to separate untrusted data from the agent's command context.
[EXTERNAL_DOWNLOADS]: The skill interacts with external services to fulfill its SEO analysis functions.
Evidence: Utilizes Firecrawl and Tavily for live search results.
Context: These are well-known, established search and SEO tools, and their use is consistent with the skill's stated purpose.

serp-analysis