SkillAgentSearch skills...

xedit-automation

Use whenever the task involves inspecting, modifying, or building plugins for Bethesda games via the forked xEdit automation daemon. Loads first; routes to the right path (MCP intent tool, MCP atomic passthrough, or sub-agent delegation) and prevents the agent from bypassing the harness.

Install / Use

npx skills add hashgraph-online/awesome-codex-plugins --skill xedit-automation

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

93/100

Category

Automation

Supported Platforms

Universal

Our assessment of xedit-automation

xedit-automation scores 93/100 on our quality scale, 715th of 2,790 Automation skills we index (top 26%).

Its SKILL.md is 26 KB long, well organised into 25 sections with 11 code examples: a thorough specification that gives an agent plenty to work with.

With 1,075 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
13/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated about 2 months ago, so xedit-automation is actively maintained.
  • Our last check on 2026-10-25 found the source still online.
  • It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-10-07. Automated pattern scan on 2026-10-07. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

xedit-automation compared with similar skills

All 4 of these similar skills score higher than xedit-automation; compare them before choosing.

SkillScoreStarsUpdatedFormat
xedit-automation (this skill)by hashgraph-online931.1k53d agoSKILL.md
Agent-Reachby Panniantong100112.3k8d agoCLAUDE.md
rufloby ruvnet10081.9ktodayCLAUDE.md
headroomby headroomlabs-ai10079.8ktodayCLAUDE.md
CowAgentby zhayujie10048.0ktodayCLAUDE.md

Frequently asked questions

How do I install xedit-automation?
Run npx skills add hashgraph-online/awesome-codex-plugins --skill xedit-automation. The install tabs above show the steps for each supported agent.
Which AI agents does xedit-automation work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is xedit-automation safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is Apache-2.0-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is xedit-automation still maintained?
The repository was last updated about 2 months ago, so xedit-automation is actively maintained.

name: xedit-automation description: Use whenever the task involves inspecting, modifying, or building plugins for Bethesda games via the forked xEdit automation daemon. Loads first; routes to the right path (MCP intent tool, MCP atomic passthrough, or sub-agent delegation) and prevents the agent from bypassing the harness.

xEdit Automation — Hub Skill

This skill is the always-loaded entry point for any xEdit work. It is the single source of truth for "which path do I use" and "what must I never do." Specialised task skills (e.g. xedit-conflict-audit) inherit its routing, anti-patterns, and verification discipline; they do not restate them.

Toolbox at a glance (progressive disclosure, r6-aware)

The forked xEdit daemon is best treated as a progressive-disclosure surface: start with the small MCP intent tools, ask the live daemon which r6 capability blocks it supports, then switch to the richer one-call patterns only when the corresponding system.capabilities.supports.* key is present.

  • Discovery & session — xedit_session, xedit_list_capabilities. Call xedit_session first every conversation. Then call xedit_list_capabilities once to see the command digest, contractVersionExpected, and r6 supports.* anchors.
  • Reading records & conflicts — xedit_find_record, xedit_read_record, xedit_inspect_conflicts. These are the W2 (conflict audit) backbone; branch to atomic passthrough when W2 needs an r6 response block that no intent tool exposes yet.
  • Atomic passthrough — xedit_call(command, args). For any native daemon command that does not have an intent tool yet. Still runs the full pipeline (validation → state → rules → audit). Use it whenever the intent tools do not fit.

For deep reference material, query the structured BGS KB first (bgs_kb_query / bgs_kb_get). Deep reference records live under knowledge/bgs-kb/packs/core/records/xedit/ (queryable via bgs_kb_query / bgs_kb_get).

Routing doctrine (which path to use)

| Task shape | Path | |---|---| | High-frequency known intent (audit a conflict, read a record, run a job, write a patch) | MCP intent tool | | Novel / debugging / free composition of native commands | MCP atomic passthrough: xedit_call(command, args) — still in harness | | Exploratory atomic-op storm (trial-and-error, repeated read-eval, hypothesis testing) | Delegate to a read-only investigator sub-agent with this skill loaded; the sub-agent burns its own context, returns a distilled summary | | Large formalisable bulk mutation | MCP xedit_run_script (Batch 4+) with dry-run + snapshot | | Daemon explicitly in default (non-MCP) mode, manual debug only | Direct xedit-client.ps1 is acceptable — but ONLY when the user has explicitly accepted the risk and the daemon is not in -automation-mcp-mode |

The agent should never have a reason to bypass the MCP. Atomic passthrough exists for that.

补丁创作判断 / Patch authoring judgment

Overview

Patch authoring is not "make the red go away." It is a small compatibility argument: preserve the current winner, forward only the upstream values that should still survive, and leave behind a reversible patch plugin whose contents say what you meant. Sorting is a single-choice lever; patching is how you keep two mods' intended values active at once. But true-in-principle is not true-at-pack-scale: if the plan is to stitch every FormID because "xEdit can fix anything," the plan is already broken.

Use this section when you are about to author or judge a patch through the MCP. Use xedit-conflict-audit first when the question is still "what wins, what conflicts, and should this be patch-vs-reorder?" That skill owns the patch-vs-reorder call; this section owns whether the patch you are about to write is well-formed.

Decision flow: is this patch well-formed?

digraph patch_authoring_judgment {
  // Pending-shutdown saves block lifecycle operations; see Dirty-state and
  // relaunch control below. A restart is not a durability mechanism.
  rankdir=TB;
  node [shape=box];

  start [shape=doublecircle, label="Need a patch?"];
  audit [label="Audit actual winner + conflict fields\n(xedit-conflict-audit if not already done)"];
  reorder [shape=diamond, label="Is ordering enough\nwithout losing needed values?"];
  no_patch [shape=doublecircle, label="No patch: reorder / leave winner\nthen verify readback"];
  winner [label="Copy/target the current winner\nnot an older loser"];
  intent [label="Name each value to forward\nWhat does it do? Why should it survive?"];
  noise [shape=diamond, label="Any unchanged copied values\nor green/ITM-style noise?"];
  clean [label="Remove unchanged/no-intent values"];
  refs [shape=diamond, label="Deleting or mark-deleting\nrecords/elements?"];
  referenced [label="Run referenced_by / reference audit\nStop if consequences are unknown"];
  esl [label="Choose patch form deliberately\nUsually ESP flagged ESL when safe; avoid native ESL/ESM if sort freedom matters"];
  reversible [shape=diamond, label="Separate reversible patch\n(no direct source-mod edit, no unmanaged overwrite spill)?"];
  fix_shape [label="Move into managed MO2 mod\nor create dedicated patch plugin"];
  verify [label="Save + restart + readback\nwinning override shows intended values"];
  pass [shape=doublecircle, label="Well-formed patch"];

  start -> audit -> reorder;
  reorder -> no_patch [label="yes"];
  reorder -> winner [label="no, need combined values"];
  winner -> intent -> noise;
  noise -> clean [label="yes"];
  noise -> refs [label="no"];
  clean -> refs;
  refs -> referenced [label="yes"];
  refs -> esl [label="no"];
  referenced -> esl;
  esl -> reversible;
  reversible -> fix_shape [label="no"];
  reversible -> verify [label="yes"];
  fix_shape -> verify -> pass;
  // Contract 0.23 pending saves use xedit_flush; restart alone is not proof.
  verify [label="Save + inspect pending state; xedit_flush if needed; fresh readback"];
}

Deferred-save correction: The diagram's historical Save + restart + readback node is superseded. If session.save reports a nonzero savePendingShutdownCount, inspect xedit_dirty, then use xedit_flush on a contract-0.23 daemon. The tool validates the drain response and confirms the daemon's promised self-exit. A partial or unknown lastFlush is not durability proof. Normal stop/restart remains refused while a live pending queue exists; force:true abandons state and is never evidence of durability.

Authoring checks

  • Forward the winner. The game reads the final/winning override for a FormID. Build the patch from the current winner, then forward only the older mod values that still need to survive.
  • No unchanged cargo. If a copied value does not change meaning, it is ITM-style noise, not compatibility. Remove it unless you can state why it is intentionally carried.
  • Patch as separate layer. Do not drag values directly into the source mod just because it is the rightmost column today. A patch should be reversible by disabling/removing the patch layer, not by hand-deleting fields from someone else's plugin.
  • Preserve references. If deletion or mark-delete enters the plan, stop and run the referenced-by/readback path first. Unknown references mean unknown blast radius.
  • Choose plugin form deliberately. Ordinary compatibility patches usually want normal ESP sort freedom plus ESL flagging when safe. Native ESL/ESM load-order behavior is not a free cleanliness upgrade.
  • Good patch vs band-aid patch. A good patch has named intent per forwarded value and reduces conflict debt. A band-aid patch copies whole records, edits source mods, or tries to stitch every FormID because sorting was treated as irrelevant.

KB query discipline

This judgment section is game-agnostic. Query KB for game-specific patching gotchas, record-family caveats, and current toolchain facts rather than fossilizing them here.

bgs_kb_query({ query: "xedit patch authoring gotchas", domains: ["xedit", "load-order"], games: ["<current game>"] })

If KB is silent, mark [GAP] and keep the patching rule at the framework level instead of inventing a game-specific doctrine.

Red Flags (STOP)

| Thought | Reality | |---|---| | "It's red, so copy the left value." | Red is a prompt to inspect. It is not a verdict that the losing value should survive. | | "The rightmost mod is wrong; I'll edit it directly." | Then you made the source mod the patch. Use a separate override patch so reversal is clean. | | "Sorting does not matter because xEdit can patch anything." | In principle, yes; across hundreds of thousands of FormIDs, no. Use ordering to reduce repair debt, then patch the meaningful remainder. | | "Green/unchanged copied values are harmless." | They are no-intent cargo. If they do not change meaning, remove them. | | "Native ESL is cleaner than ESP flagged ESL." | Native ESL/ESM can cost normal sort freedom. Use ESP flagged ESL when that is the safe patch shape. | | "This looks unused; delete it." | Referenced-by first. Deletion without reference knowledge is unknown blast radius. |

Rationalizations

| Excuse | Reality | |---|---| | "I'll just drag one field into the winning plugin." | One field is still a source-mod edit. Make a patch layer, or future-you hand-removes stray fields one by one. | | "The patch should preserve everything any mod touched." | Preserve intent, not history. Some edits are core to the mod; some are incidental author preference that should lose to a systemic rule. | | "I do not know what this field does, so I'll copy the version that looks plausible." | Unknown field meaning is a research task, not a dice roll. Read the record structure, CK/wiki/community examples, and actual in-pack behavior. | | "Overwrite is fine; MO2 sees it." | Overwrite is spill. A real patch belongs in a managed mod layer with a name future-you can understand. | | "The response said ok; the patch is done." | Patch acceptance is winning-override readback after save/restart, with the intended values visible and no unintended cargo. |

R6 progressive-disclosure capability checks

Do not assume every daemon is r6. Read xedit_list_capabilities once and branch on the support keys below. On pre-r6 daemons, fall back to the older explicit record-list / child-walk / per-record loop patterns; on r6+ daemons, use the one-call or page-aware form to reduce round-trips and preserve context.

| Capability key | Contract | Prefer this pattern | KB record | |---|---:|---|---| | supports.childGroupNavigation | 0.13 | Navigate CELL/WRLD/DIAL/QUST ChildGroups through elements.children stubs | xedit.childgroup-navigation.v1 | | supports.createParentSpec | 0.16 / 0.18 | Create records directly under parent ChildGroups with records.create parent | xedit.records-create-parent-spec.v1 | | supports.elementsChildrenPagination | 0.17 | Page elements.children with limit / offset | xedit.elements-children-pagination.v1 | | supports.reverseNavigation | 0.19 | Add includeParents:true and read relations.parents | xedit.reverse-navigation.v1 |

For the whole r6 contract delta, query KB record xedit.r6-contract-summary.v1.

ChildGroup navigation (supports.childGroupNavigation)

On r6+ daemons, elements.children on CELL, WRLD, DIAL, and QUST records returns a virtual child entry with kind: "child_group". Treat that entry as a read-only navigation stub, not as a real mutable element.

Use it one generation at a time:

elements.children({ file, formId, path: "\\Child Group" })
elements.children({ file, formId, path: "\\Child Group\\Persistent" })
elements.children({ file, formId, path: "\\Child Group\\Temporary" })
elements.children({ file, formId, path: "\\Child Group\\Visible when Distant" })
elements.children({ file, formId, path: "\\Child Group\\Block X, Y" })
elements.children(

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars1.1k
CategoryAutomation
Updated1mo ago
Forks294

Languages

Python

Trust signals

100/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

No cautions