メニュー
Test Claude Code version behaviors. Install versions, run controlled prompts, inspect session artifacts.
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
Add a deferred entry to the repo's docket — a versioned NN-{slug}.md under docket/{type}/ (tech-debt, feature, follow-up, idea, or upstream), reviewed in the normal PR flow and picked up later. Resolve by deleting the file.
Build a glanceable quest-map of a long session — pure user-intent roots, ground-truth check-marks, [from → to] ranges, PR/docket overlays, a metadata header. A shared recap-on-steroids built by a fresh unbiased agent; light at the top, full detail one cat away. On-demand orientation for when the arc of a long, compacted, multi-thread session is lost and you (and the user) need to decide what to tackle next.
Migrate the binary patcher to a new Claude Code version. Re-derive bytes for changed patches, investigate missing anchors, mark obsoleted patches, verify empirically in fresh sessions, and open a PR. Use when Claude Code releases a new version (announced or stealth-on-CDN) and the patcher needs to follow.
Absorb a new Claude Code version and/or model into this repo — the reusable recipe for what to investigate and how to judge it. Use when CC updates, the model changes, or a loud failure smells like version/model drift.
Draft a PR description into scratch/, pause for your review, then create or update the PR via gh. Gathers git context, indexes the session for semantic search, and follows the repo's CLAUDE.md PR guidelines.
Recover in-flight context from a compacted, crashed, or context-limited session. Indexes transcript for semantic search, reconstructs the session arc.
| name | version-test |
| description | Test Claude Code version behaviors. Install versions, run controlled prompts, inspect session artifacts. |
| argument-hint | <version> [target] |
| disable-model-invocation | true |
| allowed-tools | Bash, Read, Glob, Grep |
Explore Claude Code version behaviors empirically.
/version-test <version> [target]
Active: !`readlink ~/.local/bin/claude | xargs basename 2>/dev/null`
Installed: !`ls ~/.local/share/claude/versions/ 2>/dev/null | grep -E '^[0-9]' | sort -V | tr '\n' ' '`
gh api repos/anthropics/claude-code/contents/CHANGELOG.md --jq '.content' | base64 -d
Use a subagent for extensive changelog analysis to avoid context bloat.
https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
| Target | Prompt | Look for |
|---|---|---|
agent | "Launch agent to compute 1+1" | subagents/ dir vs agent files in root |
schema | Any prompt | Validate JSONL with Pydantic models |
feature | n/a | "$binary" --help output |
binary=$(.claude/skills/version-test/scripts/install-version.sh $0)
MMPP0000-0000-0000-0000-00000000000N (MM=minor, PP=patch, N=counter)
"$binary" --session-id "$SESSION_ID" -p "<prompt>"
find ~/.claude/projects -name "*$SESSION_ID*"
ls -laR "$SESSION_DIR"
For schema validation:
uv run python3 << EOF
from pydantic import TypeAdapter
import json
from claude_session.schemas.session import SessionRecord
adapter = TypeAdapter(SessionRecord)
for line in open("$SESSION_FILE"):
if line.strip():
adapter.validate_python(json.loads(line))
print("✓ valid")
EOF
claude-session delete "$SESSION_ID" --force --no-backup
| Version | Change |
|---|---|
| 2.1.2 | subagents/ directory for agent files |
| 2.1.0 | Skills, forked contexts, hooks |
| 2.0.64 | Background agents, TaskOutputTool |