seeleai/gamefinder/install-game-mods

Claude Code and agent skill by seeleai/gamefinder

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Install verified downloaded game Mods through the evidence-driven Contract V2 workflow in nexus-mods-server MCP. Use when the user explicitly asks to install one Archive plus its matching Nexus receipt or a dependency-aware Mod Bundle Manifest. Build a Dynamic Game Context, reuse a verified Method when available, otherwise research and submit a bounded Agent Proposal, freeze an immutable Install Plan, classify it as auto_safe, review_required, or blocked, and apply auto_safe Plans without an extra approval turn. Do not use for Mod discovery, downloading, manual extraction, Vortex import, Mod development, or uninstalling a committed Mod.

Install

npx vskill@latest install seeleai/gamefinder/install-game-mods

SKILL.md

Install Game Mods

Install verified material through a reviewable transaction. Keep inspection, staging, game writes, backups, verification, and recovery inside nexus-mods-server MCP.

Explicit $install-game-mods invocation is the stable production entry point. Automatic triggering is optional.

Inputs

Accept one exact Archive plus its matching .nexus-receipt.json, or one verified Bundle Manifest. Require the exact target game_root. A registered profile_id is useful but no longer mandatory: an unseen game may use an evidence-backed Dynamic Game Context.

If verified download material is missing, hand off to download-nexus-mods. Download completion supplies material but does not by itself authorize installation beyond the user's explicit request.

Read:

Core behavior

  1. Call list_game_profiles, then establish a Dynamic Game Context. Use a matching legacyProfileId; use explicit identity only for an unregistered game.
  2. Prepare an immutable Evidence Pack for each verified Archive.
  3. Query reusable installation Methods and obey proposalReadiness.recommendedAction. For select_verified_method, call instantiate_install_method with the exact candidate binding and current package unit; never reconstruct its operations by hand.
  4. Submit a Proposal only when the result permits it. Never invent a package unit. If no Method matches, research only what is needed and submit either a bounded file Proposal or one controlled-installer Proposal according to recommendedAction. Do not ask the user to develop an Adapter.
  5. Call freeze_install_plan with reviewMode: "auto_safe", then show its source, selected package, exact game root, strategy, target operations, conflicts, risk, reversibility, planId, expiry, and returned review object.
  6. Follow review.nextAction: continue in the same turn for apply_now, stop for confirmation on request_confirmation, and stop without applying on stop. Do not override the server's classification locally.
  7. Apply with only planId. For a file execution, verify the resulting legacy installationId; for a controlled installer, inspect its V2 Installation Record, static verification, observed changes, recovery state, and refreshed Game Context. Review methodLearning: a successful Agent Proposal should return a local_verified Method plus a successful Method Outcome; a reused Method should append a new Outcome without creating a Mod-specific duplicate.

For Bundles, process required dependencies in installOrder. Freeze, classify, apply, and verify one current Plan at a time; do not classify later nodes from an earlier filesystem state.

Review policy

Default every valid frozen Plan to auto_safe. Keep that classification when the Plan exactly implements the user's install request, has no unresolved choice or blocker, stays inside evidence-backed writable roots, and uses reversible transaction safeguards.

Use review_required only when at least one material condition is present:

  • the user asked to inspect the Plan before writes;
  • package, target, dependency scope, or strategy requires a user choice;
  • the Plan uses run_bundled_installer, replaces an existing file, reports high risk, or cannot prove bounded reversible side effects;
  • the Plan introduces a target or operation not reasonably implied by the explicit install request.

Use blocked for MCP blockers, protected paths, unresolved conflicts, missing capabilities, running game processes, unverifiable evidence, or recovery-required state. Never relabel a blocker as review_required or auto_safe.

After expiry, staleness, or replanning, display and classify the new Plan. Continue automatically when the new Plan is still auto_safe; never transfer an earlier classification without re-evaluation.

M2 executes bounded whole-tree and source-to-target file mappings without a prewritten Adapter. This includes manifest-free root-overlay packages, but only through exact writable paths and reviewable ensure_directory, install_new_file, or backup-backed replace_file operations. M3 additionally executes one evidence-hashed bundled installer through a fixed runtime, minimal environment, timeout, declared write roots, pre-state snapshots, backups, complete game-root side-effect observation, and bounded recovery. A controlled-installer Proposal is always high risk and must declare the smallest justified write roots and required postconditions.

Hard boundaries

Never copy, extract, execute, delete, overwrite, edit, or reconstruct game operations using shell, Python, browser automation, or generic filesystem tools.

Apply accepts only:

text
apply_agentic_install_plan(planId)

Use rollback_mod_install(transactionId) only for incomplete transaction recovery, never as uninstall. Distinguish static verification from runtime verification and never claim in-game success when runtime verification is not-run.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/seeleai/gamefinder/install-game-mods/badge)](https://verified-skill.com/skills/seeleai/gamefinder/install-game-mods)

Something wrong with this skill or its page? Tell us.