The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external markdown files from the descriptions/edges/ directory to perform audits and improvements. This represents an ingestion surface for untrusted data. The risk is mitigated by explicit documentation templates (references/edge-doc-template.md) and rigid quality rules that establish structural and behavioral boundaries for the agent. Evidence: Ingestion points are located in SKILL.md (Workflow Step 1); Boundary markers are defined in the Quality Rules section; Capability inventory includes file modification and local script execution; Sanitization is performed by the audit_edge_docs.py script which validates structure using regular expressions.\n- [COMMAND_EXECUTION]: The instructions direct the agent to run local automation scripts, such as audit_edge_docs.py, Install-CodexSkill.ps1, and Test-SchemaConsistency.ps1. These scripts are part of the intended project workflow to ensure documentation quality and schema consistency within the repository environment.

openhound-edge-docs