一键导入
llm-history-tracking
LLM history tracking guidelines. Use when updating history files or starting new feature work.
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
LLM history tracking guidelines. Use when updating history files or starting new feature work.
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 | llm-history-tracking |
| description | LLM history tracking guidelines. Use when updating history files or starting new feature work. |
| version | 1.0.0 |
This skill provides guidelines for maintaining LLM development history in the Boilerplate repo. The Cursor rule llm-history-tracking.mdc is the always-applied requirement.
Each history file must contain at most 10 sessions. When session 11 needs to be added, the file must be split.
.llm/history/active/[feature-name]/
[feature-name]-part-01.md # Sessions 1-10 (always start with part-01)
[feature-name]-part-02.md # Sessions 11-20
Always use the -part-01 suffix from the beginning, even for the first file.
When a history file has 10 sessions and you need to add session 11: create a new file with -part-02 suffix and add session 11 there. Session numbers are continuous across parts (never reset).
### Session N - YYYY-MM-DD
#### Prompt (Developer)
[Exact verbatim user prompt - never summarize]
#### Key Decisions
- Decision 1
#### Files Modified
- path/to/file.ts
#### Prompt (Developer) - Manually typed by user#### Prompt (Agent) - System-generated (e.g., clicking "Build" on a plan)Update when modifying code, config, docs, or plans. Skip for pure Q&A with no file changes.
This skill is the canonical source for history timing and entry format.
End file-modifying responses with:
LLM History: Updated .llm/history/active/[feature]/[file].md (Session N)
.cursor/rules/llm-history-tracking.mdc - Cursor rule with glob triggers.llm/LLM.md - Complete history system documentationWhen changing API routes, auth, or env-dependent behavior, add or update the corresponding integration tests and keep the test file layout consistent.
When adding or changing files under infra/k8s/ or sync targets for k8s, know what Argo CD syncs and remind the user to push to Git so the cluster can sync.
Keeps tools/generate-data seeders and docs in sync when main or management DB schema or ORM entities change. Use when adding or changing tables/entities in packages/orm, packages/management-orm, or infra database migrations.
When changing DB schema or permission dimensions, consider predefined roles and bucket_role / role-related code.
Use when adding or changing environment variables, infra/env/classification YAML, scripts/env-classification, K8s env render, or local env generation.
Common patterns for the Boilerplate HTTP API (Express)