菜单
Full documentation refresh workflow: cross-refs, narrative docs, API/diagrams, validation suite. Codifies docs/tutorials/05-docs-refresh-workflow.md. Use when refreshing project docs, auditing doc health, or preparing docs before a release.
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 | tapps-docs-refresh |
| user-invocable | true |
| description | Full documentation refresh workflow: cross-refs, narrative docs, API/diagrams, validation suite. Codifies docs/tutorials/05-docs-refresh-workflow.md. Use when refreshing project docs, auditing doc health, or preparing docs before a release. |
| allowed-tools | mcp__nlt-project-docs__docs_session_start mcp__nlt-project-docs__docs_check_cross_refs mcp__nlt-project-docs__docs_check_links mcp__nlt-project-docs__docs_generate_doc_index mcp__nlt-project-docs__docs_generate_purpose mcp__nlt-project-docs__docs_generate_onboarding mcp__nlt-project-docs__docs_generate_llms_txt mcp__nlt-project-docs__docs_generate_api mcp__nlt-project-docs__docs_generate_architecture mcp__nlt-project-docs__docs_generate_interactive_diagrams mcp__nlt-project-docs__docs_check_completeness mcp__nlt-project-docs__docs_check_freshness mcp__nlt-project-docs__docs_check_drift mcp__nlt-project-docs__docs_check_diataxis mcp__nlt-project-docs__docs_check_style |
| argument-hint | [--exclude docs/archive] |
Run the full documentation refresh pipeline. Requires nlt-project-docs (and nlt-build for code edits). Do not skip failed validation steps.
Phase 0 — Scope: Exclude docs/archive/** from validation. Tier-1 targets: PURPOSE.md, ONBOARDING.md, ARCHITECTURE.md, docs/api/*, docs/adr/*.
Phase 1 — Navigation:
mcp__nlt-project-docs__docs_check_cross_refs(doc_dirs="docs", exclude="docs/archive")mcp__nlt-project-docs__docs_check_links(broken_only=true)mcp__nlt-project-docs__docs_generate_doc_index(doc_dirs="docs,README.md,AGENTS.md", output_path="docs/INDEX.md")Phase 2 — Narrative: mcp__nlt-project-docs__docs_generate_purpose, docs_generate_onboarding, docs_generate_llms_txt(mode="compact") — hand-edit placeholders after generation.
Phase 3 — API & diagrams: Regenerate docs_generate_api per package; docs_generate_architecture; docs_generate_interactive_diagrams.
Phase 4 — Verification (targets): completeness ≥ 98, cross-refs ≥ 90; run docs_check_drift, docs_check_freshness(summary_only=true), docs_check_diataxis.
Phase 5 — Style (optional): docs_check_style on tier-1 narrative files only — skip auto-generated docs/api/*.
Report: Summary table of scores, broken links count, drift findings, files written.
Persistent cross-session memory + knowledge graph for AI coding agents. Use when you need to recall prior decisions before making a non-trivial choice, save the rationale behind a decision so it survives the session, share findings across agents in a project (Hive), or record an experience event / KG triple. Talk to the deployed brain at http://127.0.0.1:8080/mcp/ via MCP tools — never via raw psycopg.
Frontend work flow combining UX playbook and standard finish pipeline. Use when the task is primarily UI/UX implementation or accessibility.
Write a structured cross-session handoff and close the TAPPS session lifecycle so the next chat can continue without a long paste. Use when ending a session, handing off to a fresh chat, or the user says hand off, save session state, or continue next time.
Frontend work flow combining UX playbook and standard finish pipeline. Use when the task is primarily UI/UX implementation or accessibility.
Write a structured cross-session handoff and close the TAPPS session lifecycle so the next chat can continue without a long paste. Use when ending a session, handing off to a fresh chat, or the user says hand off, save session state, or continue next time.
Persistent cross-session memory + knowledge graph for AI coding agents. Use when you need to recall prior decisions before making a non-trivial choice, save the rationale behind a decision so it survives the session, share findings across agents in a project (Hive), or record an experience event / KG triple. Talk to the deployed brain at http://127.0.0.1:8080/mcp/ via MCP tools — never via raw psycopg.