The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided blog content which is subsequently handled by scoring scripts and transmitted to external CMS platforms.
Ingestion points: User-provided blog content draft and metadata.
Boundary markers: A mandatory human approval gate is enforced; the agent must present an Execution Summary and receive an explicit "yes" from the user before proceeding (detailed in the "Execution gate" section).
Capability inventory: Accesses WordPress and Webflow APIs via MCP, submits to Google Search Console, and executes several local Python scripts.
Sanitization: Not explicitly mentioned, though the mandatory human review serves as a significant oversight mechanism.
[COMMAND_EXECUTION]: The skill executes local Python scripts bundled with the plugin to manage workflows and scoring.
Evidence: Invokes python "${CLAUDE_PLUGIN_ROOT}/scripts/approval-manager.py", content-scorer.py, brand-voice-scorer.py, and execution-tracker.py. These scripts are located within the plugin's environment and are used for intended functional quality gates.

publish-blog