The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides URLs to download modding tools (ThomasJepp.SaintsRow rev133) and source code from a community-hosted domain (minimaul.saintsrowmods.com). While these are standard in the modding community, they originate from non-primary vendor infrastructure.
Evidence: URLs for ThomasJepp.SaintsRow-rev133.7z and saintsrowmods-src.zip are provided in the toolchain section of SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted game data files, specifically .xtbl (XML) and .lua files, which could theoretically contain malicious prompt instructions designed to influence the agent's behavior during modding tasks.
Ingestion points: Processes .xtbl, .lua, and .asm_pc files from game directories and user-provided mod workspaces as described in SKILL.md and references/formats.md.
Boundary markers: The skill instructions do not specify the use of delimiters or warnings to ignore embedded instructions within the processed game data.
Capability inventory: The skill has the ability to execute PowerShell scripts (Update.ps1), Python scripts (KnowledgeSearch.py, run_tests.py), and third-party executables (ThomasJepp.SaintsRow.BuildPackfile.exe). It also performs file system writes during mod building and merging.
Sanitization: While the skill mentions "static lint" for Lua and "structured edits" for XTBL, there is no evidence of specific sanitization to prevent prompt injection from data fields interpreted as natural language.

saints-row-modding