菜单
Record a manual session entry into the developer's workspace journal. Use when the user wants to write down notes between tasks (explorations, observations, status updates).
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
Extract a feature SPEC from an existing codebase for brownfield Ark adoption. Produces specs/features/<slug>/SPEC.md without faking a deep-tier task. Use when the project already implements the feature and Ark needs a SPEC to reference.
Close out an Ark task by atomically committing work + task.toml + (deep tier) SPEC in one git commit. Use when the user says they're done with a task and wants to land it. Replaces the older ark-archive skill; bulk archive is now a manager-only `ark archive` CLI.
Start a standard or deep-tier Ark task. Produces PRD then PLAN, optional REVIEW on deep tier, then EXECUTE and VERIFY. Use when the user asks for feature work, refactoring, or architectural changes.
Remove an unarchived Ark task in one step. Refuses without --force when seeded files have user content. Use when the user wants to throw away a task they no longer want.
Audit every SPEC's constraints for actuator-tag health. Reports untagged rules, malformed tags, and tag/bucket mismatches across project and feature SPECs, then offers to fix them yourself or with agent assistance.
Start a research-tier Ark task. Gather a reference corpus on a topic; follow-up implementation is optional. Use when the user wants to investigate a direction before committing to a design.
| name | ark-record |
| description | Record a manual session entry into the developer's workspace journal. Use when the user wants to write down notes between tasks (explorations, observations, status updates). |
ark-record <task description>Append a manual session entry (not tied to any task) to the developer's active journal under .ark/workspace/<dev>/journal-N.md. Use this for inter-task notes: explorations, investigations, observations.
Task-driven entries are written automatically by ark-commit; do not run ark-record for those.
.ark/.developer exists (run ark init --developer <name> to bootstrap).[AI]ark context --scope record --format json
Returns resolved identity, active_journal_path, session count, branch, journal_max_lines. Use <task description> as the title or prompt for one if absent.
[AI]Append a block to active_journal_path containing exactly three agent-authored sections:
## Session N: <title>
### Summary
<one-line summary>
### Main Changes
| Area | Description |
|------|-------------|
| <area> | <description> |
Do not write **Date**, **Slug**, **Branch** — the CLI inserts those after your ## Session N: heading. Show the user what you wrote.
[AI]ark agent workspace record --manual
The CLI: resolves identity, locates the active journal-N.md, stamps **Date**, **Slug**: -, **Branch** after your last ## Session N: heading, updates the personal Session History row, refreshes top-level Active Developers. Transactional with rollback.
| Code | Cause | Recovery |
|---|---|---|
MissingIdentity | .ark/.developer not present | Run ark init --developer <name> |
EntryFileMalformed | journal does not end with a ## Session N: <title> heading | Step 2 must place the heading at the very end, no trailing content after the Main Changes table |
JournalDriftDetected | concurrent appender wrote bytes after this transaction | Journal is left intact; investigate manually |
workflow.md §8 (state & multi-session)ark-commit — task-driven journal entries (do not use ark-record for those)