メニュー
Use after skill updates to protect, classify, merge, discard, or defer local user changes. Handles undecided pending-merges manifest entries, install backups, a six-step UX, adversarial verification, and cross-platform decision propagation.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | merge-companion |
| description | Use after skill updates to protect, classify, merge, discard, or defer local user changes. Handles undecided pending-merges manifest entries, install backups, a six-step UX, adversarial verification, and cross-platform decision propagation. |
| type | lifecycle-gate |
| compatibility | ["Python 3.11+ standard library"] |
merge-companion protects local user changes during skill updates. The installer
backs up changed installed files into
~/.ghost-alice/pending-merges/<platform>/ and records them in a manifest. At
session start, the platform hook checks the manifest and surfaces this skill
when undecided entries exist.
~/.ghost-alice/pending-merges/<current-platform>/manifest.json exists.decided=false.Use the manifest-missing-or-invalid silent-pass rule:
entries emptyAll three pass without user-facing noise.
This skill is expected to surface if any of these layers are alive:
install_hooks.py~/.codex/AGENTS.mdinstall.sh / install.ps1 tail message~/.ghost-alice/pending-merges/<platform>/READ-ME-FIRST.md|| trueThis eight-layer list is the current operating contract.
references/impact-tier.md. This is display sorting only;
it does not bypass verification.later is the default and keeps the
entry pending. Advanced flows may offer AI partial merge.~/.ghost-alice/staged/<entry-id>/. Do not touch the live file yet.All AI output in steps 3, 4, and 5 must pass adversarial-verification before it is shown as a decision aid or applied. Do not skip this for cost reasons or for low-tier changes. Block immediately the moment a rationalization to skip begins.
source_path in other platform manifests.This skill may be surfaced only by:
Other skills must not call it speculatively.
Record manifest reads/writes, backup reads/copies, staged writes, and user
confirmation prompts in the final [io-trace] block.
references/manifest-schema.mdreferences/impact-tier.mdreferences/merge-flow.mdUpdate the per-session intent ledger for every user input and provide current intent as semantic context consumed by gates, skill-evolution, and jailbreak-detector. Keep intent summaries without storing raw prompts.
Use when the user asks about skill upgrades, update recommendations, or an evolution backlog; analyze io-trace and conduct_feedback to produce report-only evolution candidates. No automatic edits.
Use when an implementation plan already exists. Execute it in a separate session with review checkpoints.
Use when implementing a known feature, bug fix, or refactor with a resolved behavior contract. Requires a failing test before production code; use discovery first when the target or premise is unknown.
Use at the start of every coding or development conversation. Defines how to find and invoke coding-convention family skills. Requires relevant skills before any response, including clarification questions.
Runs after session-intent-analyzer and jailbreak-detector/downstream gate as their consumer. Decomposes the request and routes output, verification, lifecycle, and boundary skills without owning intake, raw intent inference, or tool permission.