socaseinpoint

Apply to the next suitable job from your queue. Picks the next pending role from job_search.md, verifies the posting is live, fetches the JD, presents the link and how to apply — then (after user confirms) adapts the CV and generates a PDF. Updates status and source stats after user confirms they applied.

Regenerate your CV. Knows your full profile, stack, and experience. Adapts the CV for a specific vacancy or context if provided, or refreshes the base version. Invoke when you ask to update, adapt, or regenerate the CV.

Find new remote job vacancies matching your profile. Scans Telegram channels and web sources. Filters by your stack, timezone, and pay requirements. Tags each job with its source. Updates job_search.md with new matches and source stats.

Разобрать темку по-взрослому. Берёт идею или название темки, оценивает рынок, конкурентов, входную точку, риски и выдаёт честный вердикт. Вызывай когда хочешь понять стоит ли тратить время.

Найти новые темки под твой профиль. Сканирует Telegram-каналы и веб-источники, фильтрует по стеку и интересам, обновляет temki.md с тегами источников и статистикой.

Обновить статус темки, посмотреть очередь, проверить что в работе. Вызывай когда хочешь отметить прогресс, закрыть темку или посмотреть общую картину.

Close out the CURRENT session in place — compress what you did THIS session into one verified orca Trail block, without opening a new window. The mirror of orca-resume (which folds a PRIOR session at start); flush folds the live current session as a deliberate end-of-work act. Dispatches a subagent that reads THIS session's transcript + git since the last anchor, produces one 4-layer block (done with anchors / why / next / head), appends it to the in-work focus's

Resume a project with trusted, owned session continuity. On sit-down at session start, shows the in-work focuses and forks — continue an in-work focus, finish a done one and start fresh, or start new. On "continue" it dispatches a subagent that reads the PRIOR session transcript + git log and compresses them into ONE verified 4-layer block (done with anchors / why / next / head), appends it to the focus's

Channel-selectable research — pick where to search (web, youtube, your own files) and how deep, fan out a subagent per source, adversarially verify each claim, synthesize a cited answer. The run is self-contained in an arc (question + answer together); promotion to a durable knowledge doc is optional. Invoke on "наресёрчь", "research X", "/lore", "изучи тему", "build lore about X", "есть ли тула под".

Activate whenever the current project has a `.arcs/` directory. That directory is the user's opt-in that work here runs with the arcs method — so you MUST structure non-trivial work through arcs, not just acknowledge the dir. Before doing the task, create the arc/goal and work inside it (input→workspace→output); the user does NOT learn the CLI — you run the `arcs` commands. Also activate on "arcs", "arc", "goal", "заведи арку", "arcs status". No `.arcs/`? The user has not opted in — don't create it unless they run `arcs init` or ask.