本文へ移動
cccskills
無料GitHub で公開

gsd-reapply-patches

Use after a GSD update, when local modifications were backed up to gsd-local-patches/ and must be merged back. Three-way compares the pristine baseline, the backup, and the new install, flagging conflicts instead of skipping files.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md11.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

<purpose> After a GSD update wipes and reinstalls files, this command merges user's previously saved local modifications back into the new version. Uses three-way comparison (pristine baseline, user-modified backup, newly installed version) to reliably distinguish user customizations from version drift.

Critical invariant: Every file in gsd-local-patches/ was backed up because the installer's hash comparison detected it was modified. The workflow must NEVER conclude "no custom content" for any backed-up file — that is a logical contradiction. When in doubt, classify as CONFLICT requiring user review, not SKIP. </purpose>

<process>

Step 1: Detect backed-up patches

Check for local patches directory:

expand_home() {
  case "$1" in
    "~/"*) printf '%s/%s\n' "$HOME" "${1#~/}" ;;
    *) printf '%s\n' "$1" ;;
  esac
}

PATCHES_DIR=""

# Env overrides first — covers custom config directories used with --config-dir
if [ -n "$KILO_CONFIG_DIR" ]; then
  candidate="$(expand_home "$KILO_CONFIG_DIR")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
elif [ -n "$KILO_CONFIG" ]; then
  candidate="$(dirname "$(expand_home "$KILO_CONFIG")")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
elif [ -n "$XDG_CONFIG_HOME" ]; then
  candidate="$(expand_home "$XDG_CONFIG_HOME")/kilo/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
fi

if [ -z "$PATCHES_DIR" ] && [ -n "$OPENCODE_CONFIG_DIR" ]; then
  candidate="$(expand_home "$OPENCODE_CONFIG_DIR")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
elif [ -z "$PATCHES_DIR" ] && [ -n "$OPENCODE_CONFIG" ]; then
  candidate="$(dirname "$(expand_home "$OPENCODE_CONFIG")")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
elif [ -z "$PATCHES_DIR" ] && [ -n "$XDG_CONFIG_HOME" ]; then
  candidate="$(expand_home "$XDG_CONFIG_HOME")/opencode/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
fi

if [ -z "$PATCHES_DIR" ] && [ -n "$GEMINI_CONFIG_DIR" ]; then
  candidate="$(expand_home "$GEMINI_CONFIG_DIR")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
fi

if [ -z "$PATCHES_DIR" ] && [ -n "$CODEX_HOME" ]; then
  candidate="$(expand_home "$CODEX_HOME")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
fi

if [ -z "$PATCHES_DIR" ] && [ -n "$CLAUDE_CONFIG_DIR" ]; then
  candidate="$(expand_home "$CLAUDE_CONFIG_DIR")/gsd-local-patches"
  if [ -d "$candidate" ]; then
    PATCHES_DIR="$candidate"
  fi
fi

# Global install — detect runtime config directory defaults
if [ -z "$PATCHES_DIR" ]; then
  if [ -d "$HOME/.config/kilo/gsd-local-patches" ]; then
    PATCHES_DIR="$HOME/.config/kilo/gsd-local-patches"
  elif [ -d "$HOME/.config/opencode/gsd-local-patches" ]; then
    PATCHES_DIR="$HOME/.config/opencode/gsd-local-patches"
  elif [ -d "$HOME/.opencode/gsd-local-patches" ]; then
    PATCHES_DIR="$HOME/.opencode/gsd-local-patches"
  elif [ -d "$HOME/.gemini/gsd-local-patches" ]; then
    PATCHES_DIR="$HOME/.gemini/gsd-local-patches"
  elif [ -d "$HOME/.codex/gsd-local-patches" ]; then
    PATCHES_DIR="$HOME/.codex/gsd-local-patches"
  else
    PATCHES_DIR="$HOME/.claude/gsd-local-patches"
  fi
fi
# Local install fallback — check all runtime directories
if [ ! -d "$PATCHES_DIR" ]; then
  for dir in .config/kilo .kilo .config/opencode .opencode .gemini .codex .claude; do
    if [ -d "./$dir/gsd-local-patches" ]; then
      PATCHES_DIR="./$dir/gsd-local-patches"
      break
    fi
  done
fi

Read backup-meta.json from the patches directory.

If no patches found:

No local patches found. Nothing to reapply.

Local patches are automatically saved when you run /gsd-update
after modifying any GSD workflow, command, or agent files.

Exit.

Step 2: Determine baseline for three-way comparison

The quality of the merge depends on having a pristine baseline — the original unmodified version of each file from the pre-update GSD release. This enables three-way comparison:

  • Pristine baseline (original GSD file before any user edits)
  • User's version (backed up in gsd-local-patches/)
  • New version (freshly installed after update)

Check for baseline sources in priority order:

Option A: Git history (most reliable)

If the config directory is a git repository:

CONFIG_DIR=$(dirname "$PATCHES_DIR")
if git -C "$CONFIG_DIR" rev-parse --git-dir >/dev/null 2>&1; then
  HAS_GIT=true
fi

When HAS_GIT=true, use git log to find the commit where GSD was originally installed (before user edits). For each file, the pristine baseline can be extracted with:

git -C "$CONFIG_DIR" log --diff-filter=A --format="%H" -- "{file_path}"

This gives the commit that first added the file (the install commit). Extract the pristine version:

git -C "$CONFIG_DIR" show {install_commit}:{file_path}

Option B: Pristine snapshot directory

Check if a gsd-pristine/ directory exists alongside gsd-local-patches/:

PRISTINE_DIR="$CONFIG_DIR/gsd-pristine"

If it exists, the installer saved pristine copies at install time. Use these as the baseline.

Option C: No baseline available (two-way fallback)

If neither git history nor pristine snapshots are available, fall back to two-way comparison — but with strengthened heuristics (see Step 3).

Step 3: Show patch summary

## Local Patches to Reapply

**Backed up from:** v{from_version}
**Current version:** {read VERSION file}
**Files modified:** {count}
**Merge strategy:** {three-way (git) | three-way (pristine) | two-way (enhanced)}

| # | File | Status |
|---|------|--------|
| 1 | {file_path} | Pending |
| 2 | {file_path} | Pending |

Step 4: Merge each file

For each file in backup-meta.json:

  1. Read the backed-up version (user's modified copy from gsd-local-patches/)
  2. Read the newly installed version (current file after update)
  3. If available, read the pristine baseline (from git history or gsd-pristine/)

Three-way merge (when baseline is available)

Compare the three versions to isolate changes:

  • User changes = diff(pristine → user's version) — these are the customizations to preserve
  • Upstream changes = diff(pristine → new version) — these are version updates to accept

Merge rules:

  • Sections changed only by user → apply user's version
  • Sections changed only by upstream → accept upstream version
  • Sections changed by both → flag as CONFLICT, show both, ask user
  • Sections unchanged by either → use new version (identical to all three)

Two-way merge (fallback when no baseline)

When no pristine baseline is available, use these strengthened heuristics:

CRITICAL RULE: Every file in this backup directory was explicitly detected as modified by the installer's SHA-256 hash comparison. "No custom content" is never a valid conclusion.

For each file: a. Read both versions completely b. Identify ALL differences, then classify each as:

  • Mechanical drift — path substitutions (e.g. /Users/xxx/.claude/ → $HOME/.claude/), variable additions (${GSD_WS}, ${AGENT_SKILLS_*}), error handling additions (|| true)
  • User customization — added steps/sections, removed sections, reordered content, changed behavior, added frontmatter fields, modified instructions

c. If ANY differences remain after filtering out mechanical drift → those are user customizations. Merge them. d. If ALL differences appear to be mechanical drift → still flag as CONFLICT. The installer's hash check already proved this file was modified. Ask the user: "This file appears to only have path/variable differences. Were there intentional customizations?" Do NOT silently skip.

Git-enhanced two-way merge

When the config directory is a git repo but the pristine install commit can't be found, use commit history to identify user changes:

# Find non-update commits that touched this file
git -C "$CONFIG_DIR" log --oneline --no-merges -- "{file_path}" | grep -v "gsd:update\|GSD update\|gsd-install"

Each matching commit represents an intentional user modification. Use the commit messages and diffs to understand what was changed and why.

  1. Write merged result to the installed location

Post-merge verification

After writing each merged file, verify that user modifications survived the merge:

  1. Line-count check: Count lines in the backup and the merged result. If the merged result has fewer lines than the backup minus the expected upstream removals, flag for review.

  2. Hunk presence check: For each user-added section identified during diff analysis, search the merged output for at least the first significant line (non-blank, non-comment) of each addition. Missing signature lines indicate a dropped hunk.

  3. Report warnings inline (do not block):

    ⚠ Potential dropped content in {file_path}:
      - Missing hunk near line {N}: "{first_line_preview}..." ({line_count} lines)
      - Backup available: {patches_dir}/{file_path}
    
  4. Track verification status — add to per-file report: Merged (verified) vs Merged (⚠ {N} hunks may be missing)

  5. Report status per file:

    • Merged — user modifications applied cleanly (show summary of what was preserved)
    • Conflict — user reviewed and chose resolution
    • Incorporated — user's modification was already adopted upstream (only valid when pristine baseline confirms this)

Never report Skipped — no custom content. If a file is in the backup, it has custom content.

Step 5: Cleanup option

Ask user:

  • "Keep patch backups for reference?" → preserve gsd-local-patches/
  • "Clean up patch backups?" → remove gsd-local-patches/ directory

Step 6: Report

## Patches Reapplied

| # | File | Result | User Changes Preserved |
|---|------|--------|----------------------|
| 1 | {file_path} | Merged | Added step X, modified section Y |
| 2 | {file_path} | Incorporated | Already in upstream v{version} |
| 3 | {file_path} | Conflict resolved | User chose: keep custom section |

{count} file(s) updated. Your local modifications are active again.
</process>

<success_criteria>

  • All backed-up patches processed — zero files left unhandled
  • No file classified as "no custom content" or "SKIP" — every backed-up file is definitionally modified
  • Three-way merge used when pristine baseline available (git history or gsd-pristine/)
  • User modifications identified and merged into new version
  • Conflicts surfaced to user with both versions shown
  • Status reported for each file with summary of what was preserved
  • Post-merge verification checks each file for dropped hunks and warns if content appears missing </success_criteria>

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Train and optimize AI agents using Microsoft's Agent Lightning framework with reinforcement learning. Use when setting up agent training, instrumenting agents with tracing, configuring LightningStore, implementing reward functions, or optimizing prompts with RL/APO algorithms.

日本語の概要は準備中です。原文の説明を表示しています。

coco-research/coco5852026年10月13日 更新

Post-run self-evaluation system that scores agent output on correctness, clarity, actionability, and conciseness. Use after /team runs, skill executions, or when explicitly asked to evaluate output quality.

日本語の概要は準備中です。原文の説明を表示しています。

coco-research/coco5852026年10月13日 更新

Create AI marketing videos for ads, promos, product launches, and brand content. Models: Veo, Seedance, Wan, FLUX for visuals, Kokoro for voiceover. Types: product demos, testimonials, explainers, social ads, brand videos. Use for: Facebook ads, YouTube ads, product launches, brand awareness. Triggers: marketing video, ad video, promo video, commercial, brand video, product video, explainer video, ad creative, video ad, facebook ad video, youtube ad, instagram ad, tiktok ad, promotional video, launch video

日本語の概要は準備中です。原文の説明を表示しています。

coco-research/coco5852026年10月13日 更新

Use when building AI features into a product: LLM integration, RAG pipelines, guardrails, streaming, AI UX, prompt engineering, or AI cost control. Treats prompts as code and validates every model output.

日本語の概要は準備中です。原文の説明を表示しています。

coco-research/coco5852026年10月13日 更新

Your AI research and engineering brain trust. 59 named personas across 8 cells covering frontier labs, applied product, model architecture, reasoning/RL/agents, alignment and interpretability, theory and science of DL, multimodal and…

日本語の概要は準備中です。原文の説明を表示しています。

coco-research/coco5852026年10月13日 更新

Use when designing a new REST or GraphQL API, reviewing an API spec before implementation, setting team API standards, or migrating REST to GraphQL. Covers resources, HTTP semantics, pagination, error handling, and pitfalls.

日本語の概要は準備中です。原文の説明を表示しています。

coco-research/coco5852026年10月13日 更新

coco-research のスキルをすべて見る

このスキルの問題を報告する