一键导入
journal
Write a journal entry for today's work session on the current project
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Write a journal entry for today's work session on the current project
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | journal |
| description | Write a journal entry for today's work session on the current project |
| disable-model-invocation | true |
Write a journal entry for today's work session on the current project.
readlink journals 2>/dev/null || echo "NOT_SYMLINKED"
git -C journals rev-parse --is-inside-work-tree 2>/dev/null || echo "NOT_GIT_REPO"
If the symlink check says NOT_SYMLINKED → STOP. Do not write anything. Tell the user: "journals/ is not a symlinked directory. Set it up first: ln -s ~/projects/work-journals/<project-name> journals"
If the git repo check says NOT_GIT_REPO → STOP. Tell the user the target is not a git repo and journalling will not be backed up.
Never commit or push anything inside journals/ to the project repo. Journals are backed up only via the separate git repo the symlink points to.
Ask the user: "Anything specific you want captured or emphasized in this entry?" — wait for a response before proceeding. If they say no or nothing specific, write from conversation context.
ls journals/*-$(date +%Y-%m-%d)-journal.md 2>/dev/null || echo "NO_FILE_FOR_TODAY"
You MUST run this command and read its output. Do NOT skip this step. Do NOT assume no file exists.--- separator at the bottom. Use ## Session N heading (increment from the last session number in the file). Add --- at the end of your entry. NEVER create a new file when one already exists for today.NO_FILE_FOR_TODAY: Create a new file with the next sequence number: <NNN>-<YYYY-MM-DD>-journal.md. Add --- at the end.# Journal Entry <NNN> - <date> - <short title>
(or ## Session 2/3/etc if appending to existing file)
## Goal
<What was the objective this session?>
## What Changed
<Concrete list of what was built/fixed/refactored>
## Key Insights
<Technical learnings, gotchas discovered, things that failed and why, things that succeeded>
## Decisions Made
<Any architectural or design decisions and their rationale>
---
--- separator for future appendingAfter writing the journal entry, also update:
CLAUDE.md — Update any new architecture patterns, known issues, or context that future sessions need.
Commit and push the journal repo — After writing the journal file, commit and push it in the work-journals repo:
cd $(readlink journals) && git add -A && git commit -m "journal: <NNN> - <YYYY-MM-DD> - <short title>" && git push
Use the symlink target path (from readlink journals) as the working directory. Never run these git commands from inside the project repo.
Auto memory MEMORY.md — Located at ~/.claude/projects/<project-path>/memory/MEMORY.md. Apply these rules strictly:
Turn a YouTube video into a markdown note in the user's notes vault — a note good enough to replace watching the video. Use when given a YouTube link to save, summarise, take notes on, or "make a note from", even if the user doesn't say "note" — any YouTube link with intent to capture it. Pulls the transcript in the video's original language, distils it with timestamps, searches the vault for notes the video connects to, writes a personalized "why this matters" section, and gives an honest watch/skip verdict.
Bootstrap context for a new session on the current project. Run this BEFORE starting any implementation work.
Siddhartha's Flutter app conventions and architecture patterns. Use when writing Flutter/Dart app code, scaffolding a new Flutter feature or whole project, designing state management, networking, offline-first sync, or dependency injection, or making architecture decisions in a Flutter project. Also use when reviewing Flutter PRs, structuring packages, or when someone asks about Flutter project conventions. Ensures consistency across Flutter projects and shared work.
Research a topic by reading many YouTube transcripts instead of watching the videos. Use when the user wants to research, learn about, or gather insights on a topic whose knowledge lives in talks, teardowns, interviews, or tutorials — product design, behavioural science, competitive/market analysis, marketing tactics, how-tos — even when they don't say "YouTube" or "transcript". A coordinator finds and curates videos, then fans out subagents to read transcripts in parallel and return cited findings. Best when watching several long videos by hand isn't worth the time.
Research a topic by reading many YouTube transcripts instead of watching the videos, using gog (YouTube Data API) for search. Use when the user wants to research, learn about, or gather insights on a topic whose knowledge lives in talks, teardowns, interviews, or tutorials — product design, behavioural science, competitive/market analysis, marketing tactics, how-tos — even when they don't say "YouTube" or "transcript", and has gog installed and prefers the YouTube Data API's relevance ranking for search. A coordinator finds and curates videos, then fans out subagents to read transcripts in parallel and return cited findings.
Siddhartha's Go backend coding conventions and architecture patterns. Use when writing Go backend code, designing APIs, setting up new Go services, or making architecture decisions in Go projects. Also use when reviewing Go PRs, scaffolding new packages, or when someone asks about project structure conventions. Also use when parallelizing work, designing concurrency (worker pools, fan-out/fan-in, rate limiting, caching, graceful degradation), or reviewing goroutine-heavy Go for races and leaks.