القائمة
Resume from a previous interrupted Claude Code session. Checks conversation logs, git status, and pending work to reconstruct context.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | resume-session |
| description | Resume from a previous interrupted Claude Code session. Checks conversation logs, git status, and pending work to reconstruct context. |
| allowed-tools | ["Bash(readonly)","Read","Glob","Grep","Task"] |
Recover context from a previous interrupted Claude Code session and identify pending tasks.
Claude Code stores conversation logs as JSONL files in the project-specific directory under ~/.claude/projects/. Determine the correct path by checking the current working directory mapping:
# The project dir name is derived from the absolute working directory with slashes replaced by dashes
ls -lt ~/.claude/projects/*/ # find the matching project directory
List the most recent log files by modification time:
ls -lt <project-log-dir>/*.jsonl | head -6
The current session is the most recently modified file. The second file is most likely the interrupted session.
Use an Explore agent (via Task tool) to read the end of the interrupted session log and extract:
"role":"human")Prompt the agent with: Read the last ~200 lines of the JSONL file. Extract user messages, active todos, and the final state of work. Identify what was incomplete when the session ended.
In parallel with the Explore agent, check:
Git status: uncommitted changes, staged files
git status
Git diff summary: what files were modified
git diff --stat
git diff --cached --stat
Recent commits: to understand what was already committed
git log --oneline -10
Auto memory: check MEMORY.md for recently documented work
Combine findings from the log analysis and git state to produce a summary:
Present this as a concise summary to the user, then ask what they'd like to prioritize.
Express coordinate reference frames, time scales, and state representations in the type system, and test them with property-based tests — durable design guidance for orts/arika that holds whether you extend the current types or redesign them. Use when implementing, refactoring, or rethinking frame/scale code: adding a frame, scale, or transform; typing vectors, epochs, covariances, or rotations; deciding which time scale a transform takes; isolating EOP dependencies; choosing an interpolation space; or writing tests for rotation and scale-conversion invariants — and also when reviewing or verifying frame/scale code for type-safety. Companion to `coordinate-time-systems`, which covers the domain concerns (which frame/scale/representation, the traps).
Reason about coordinate reference frames and time scales correctly (any astrodynamics work) — which frame/scale/representation a value is actually in, the concerns to resolve before trusting it, and what to look up. Use when interpreting or choosing frames (ECI/ECEF, GCRS/ITRS/CIRS/TIRS, J2000/EME2000, TEME, ICRF, LVLH/RTN/RSW, body-fixed), time scales (TAI/UTC/UT1/TT/TDB/GPS), or state representations (Cartesian/Keplerian/equinoctial); or when reasoning about frame/scale conversions, covariance transforms, interpolation, leap seconds, ERA/GMST/GAST, EOP/ΔUT1, polar motion, frame bias, or realizations — and also when reviewing or verifying frame/time/coordinate work (use it as a checklist of the traps, even for code you could write unaided). The companion skill `coordinate-time-implementation` covers expressing and testing these in orts/arika code.
Run an external AI code review to get independent feedback on code changes. Use this skill when the user asks for a code review, says "コードレビューして", "レビュー", "変更を見て", "code review", or wants external feedback on their recent commits or uncommitted changes.
A knowledgeable neighbor who offers a fresh perspective on architecture, design decisions, and test strategy. Use proactively when a second opinion would improve quality, or when the user says "別の視点が欲しい", "設計を相談したい", "相談", "smart-friend に聞いて", or asks for feedback on their approach.
Reference for generating Orekit cross-validation fixtures. Covers orekit-jpype setup, force model matching, CSSI space weather, and fixture generation patterns.
Guidelines for E2E testing the orts viewer with Playwright. Covers DuckDB query-level data verification, canvas pixel analysis, and common pitfalls.