一键导入
compound-v-persist
Resolves target repository and storage location for conversation artifacts. Guarantees organized, time-sorted history.
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
Resolves target repository and storage location for conversation artifacts. Guarantees organized, time-sorted history.
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 | compound-v-persist |
| description | Resolves target repository and storage location for conversation artifacts. Guarantees organized, time-sorted history. |
Use this skill BEFORE writing any artifact to .promptherder/.
When multiple repositories are open in the workspace, all .promptherder/ paths target the repository the user is working in, not the methodology source repo. Infer the target from the user's active document or recent conversation context. If the active document is outside all repositories, use recent conversation context (which files were read/written). If still ambiguous, ask which repository before writing.
Install paths: promptherder writes agent files to .agents/ by default. Antigravity currently uses .agents/ and still reads the legacy .agent/ path for backward compatibility. Global skills live at ~/.gemini/antigravity/skills/. Workspace skills live at .agents/skills/, with legacy setups using .agent/skills/.
The "slug" is the folder name in .promptherder/convos/. It MUST follow the format:
YYYY-MM-DD-kebab-case-topic
Logic flow:
YYYY-MM-DD-, prepend today's date.plan.md?ideas.md (brainstorming)?
-> CREATE NEW: Generate a short, descriptive kebab-case slug based on the goal.review-*.md?debug.md?task.md?
-> REUSE LATEST: Find the most recently modified folder in .promptherder/convos/.
plan.md, treat as New Topic..promptherder/convos/<YYYY-MM-DD>-<slug>/Return the full absolute path for the file to be written (e.g., .../2024-01-01-fix-login/plan.md).
Autonomous planning with internal reasoning. Researches, evaluates ideas, presents a plan. Minimizes user round-trips. Use before making non-trivial changes.
Reviews changes for correctness, edge cases, style, security, and maintainability with severity levels. 10 parallel checks with version-specific research. Use before finalizing changes.
Systematic debugging — reproduce, isolate, form hypotheses, instrument, fix, and add regression tests. Use when troubleshooting errors, failing tests, or unexpected behavior.
Analyzes task dependencies and groups independent steps into parallel batches. Use when executing multi-step plans or performing research across multiple sources.
Applies tests-first discipline (red/green/refactor) and adds regression tests for bugs. Use when implementing features, fixing bugs, or refactoring.
Mandatory checklist before claiming a task is done. Ensures verification, clean code, and accurate reporting. Use before saying "done" or "complete".