The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute the Loom Skill Orchestrator via dotnet so.dll or platform-specific binaries (so.exe). This execution is governed by a strict CLI contract and limited to specific, documented subcommands (compile, run, resume, status, inspect-workflow, inspect-events).
[EXTERNAL_DOWNLOADS]: The restore-so-runtime.ps1 script automates the acquisition of the orchestrator runtime from the official NuGet registry (api.nuget.org) and the author's GitHub releases (github.com/waynebaby/Techne-Loom). The script performs mandatory SHA-512 integrity verification and nuspec identity checks before allowing execution, significantly mitigating supply chain risks.
[DYNAMIC_EXECUTION]: The skill uses dotnet exec to run the orchestrator DLL at runtime. This dynamic execution is the core purpose of the skill and is protected by a "Runtime Mode Separation" policy that ensures independent paths for self-contained and .NET CLI modes, preventing cross-contamination or unauthorized fallbacks.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests and enhances target skill deliverables. However, it implements a "Boundary Check And Approval Gate" which forces the agent onto a governed route, requiring explicit machine-readable predicates or human approval for all state transitions, thereby preventing the autonomous execution of injected instructions.

loom-skill-enhancement