القائمة
Bootstrap context for a new session on the current project. Run this BEFORE starting any implementation work.
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 | prep-for-task |
| description | Bootstrap context for a new session on the current project. Run this BEFORE starting any implementation work. |
| disable-model-invocation | true |
Bootstrap context for a new session on the current project. Run this BEFORE starting any implementation work.
cat CLAUDE.md 2>/dev/null || echo "No CLAUDE.md found"
PROJECT_DIR=$(pwd | sed 's|/|-|g; s|^-||'); cat ~/.claude/projects/"$PROJECT_DIR"/memory/MEMORY.md 2>/dev/null || echo "No MEMORY.md found"
The project instructions and auto memory above have been pre-loaded. Now launch these exploration subagents in parallel to gain comprehensive project context:
Read these files and summarize the project state:
README.md — services, configuration, credentialsgit log --oneline -20 for recent commit historygit status for working tree state~/.claude/plans/*.md — read any plan files modified in the last 7 daysExplore the full directory tree and map out the project structure. Adapt based on what you find:
Read ALL files in journals/ sorted by name (ascending), if the directory exists. For each session entry, extract signal and condense it. Return a structured summary to the main instance — do NOT return raw journal content:
## Journal Summary
### Recently Built / Fixed
<bullet per notable feature or fix, most recent first>
### Technical Gotchas
<bullet per confirmed gotcha or non-obvious behaviour>
### Architectural Decisions
<bullet per decision with rationale>
### Deferred / Known Issues
<bullet per thing explicitly deferred or known to be broken>
Omit anything already superseded by a later journal entry. Be terse — this summary must fit in ~30 lines. If no journals/ directory exists, skip this agent.
Summarize your findings to the user in this format:
## Project State
- Recent features: <list>
- Known issues: <list>
## Recent Changes
<last 5 commits, one line each>
## Key Gotchas to Remember
<top 5 technical pitfalls from journals and memory>
## Suggested Next Task
Based on open issues, in-progress branches, plan files, and deferred items from journals — suggest the single most logical next task and ask if that's the priority or if something else takes precedence.
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.
Write a journal entry for today's work session on the current project
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.