The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local Python scripts to perform validation and project management tasks.
Scripts include validate_query_corpus.py, manage_markdown_site.py, and manage_markdown_work.py.
Commands are executed with parameters like --bundle, --inventory, and --output to process research artifacts.
[DYNAMIC_EXECUTION]: Utility scripts are accessed via a computed root path resolved at runtime.
The <suite-root> path is determined from environment variables such as ${CLAUDE_PLUGIN_ROOT} or values within a local configuration file named .seo-suite-runtime.json.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and classifying data from external sources, which introduces a surface for indirect injection.
Ingestion points: External data enters the system through engine answer captures, competitor result observations, and cited URL landscapes as documented in SKILL.md and references/capture-protocol.md.
Boundary markers: While no explicit prompt delimiters are provided for the ingested data, the skill uses SHA-256 hash-pinning for all raw evidence files to ensure data integrity.
Capability inventory: The skill has the capability to execute shell commands via the suite's Python scripts and write structured research artifacts to the local filesystem.
Sanitization: The references/capture-protocol.md mandates the redaction of secrets, personal data, account identifiers, and sensitive prompt content from raw source excerpts before they are stored or processed.

ai-search-research