The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python utility scripts/entity_panel_tool.py for inventorying repository assets, assembling panel fragments, and performing structural validation. It also references a PowerShell script Install-CodexSkill.ps1 for environment setup.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect injection surface as it ingests and transforms data from repository files (schema, Markdown descriptions, and JSON saved searches) into generated output. This is addressed by a mandatory evidence chain:
Ingestion points: Files located in extension/schema.json, descriptions/, and extension/saved_searches/ are read to derive panel content.
Boundary markers: The workflow uses isolated fragment files and a strict sub-agent contract (references/entity-agent-contract.md) to scope individual entity tasks.
Capability inventory: The skill uses local scripts to read/write JSON and performs Markdown/Cypher generation.
Sanitization: The included entity_panel_tool.py script provides sanitization logic that detects and rejects unsafe Markdown (e.g., javascript:, vbscript:, or data: URI schemes), raw HTML, and malformed Cypher queries.
[DYNAMIC_EXECUTION]: The skill generates Go template strings ({{ ... }}) that are designed to be evaluated at runtime by the target BloodHound platform for dynamic data exploration. Additionally, the provided test suite (tests/test_entity_panel_tool.py) utilizes exec_module to dynamically load the tool script for unit testing purposes.

openhound-entity-panels