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。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/rlaope/bestwork-agent --skill docs命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Evidence-driven causal tracing — competing hypotheses, falsification, rebuttal round, recommended next probes
Feature validation gate — research, define purpose, stress-test from real user perspective before building
Maximum parallelism burst — all tasks execute simultaneously with no coordination overhead
正在显示 SKILL.md
基于 SOC 职业分类
| name | docs |
| description | Sync all documentation (README, docs/, CLAUDE.md) with current codebase state |
When this skill is invoked, IMMEDIATELY print:
[BW] syncing docs with codebase...
Audits and updates ALL project documentation to match the current code. This is NOT a simple copy — it reads the actual codebase to generate accurate docs.
Read these files to understand what exists NOW:
package.json — version, scripts, dependencies, bin.claude-plugin/plugin.json — plugin name, versionskills/*/SKILL.md — all skill names + descriptionsprompts/tech/*.md, prompts/pm/*.md, prompts/critic/*.md — count agents per rolesrc/harness/orchestrator.ts — team modes, domain keywords, classification logicsrc/harness/smart-gateway.ts — skill routes, gateway flowhooks/hooks.json — registered hookshooks/*.sh, hooks/*.mjs — hook scripts that exist.github/workflows/*.yml — CI/CD setupCLAUDE.md — current project instructionsRun:
npm test 2>&1 | tail -3 # get test count
ls skills/ # get skill list
ls prompts/tech/ prompts/pm/ prompts/critic/ | wc -l # get agent count
Read current README.md, then update with accurate numbers and features:
Keep the existing README structure and style. Only update facts that changed. Do NOT rewrite from scratch — preserve the tone and formatting.
Same updates as README.md but in Korean/Japanese respectively. If these files exist, update them. If not, skip.
Read current file, then update:
bestwork-agent@bestwork-tools)/bestwork-agent:update or claude plugin update bestwork-agent@bestwork-tools)Read current file, then regenerate the skill catalog from the actual skills/*/SKILL.md files. For each skill, extract name and description from frontmatter.
Read current file. Update:
After all updates, run these checks:
7a. Version numbers — no stale version references:
grep -c "0.10.0\|0.9.0\|0.8.0" README.md README.ko.md README.ja.md docs/*.md CLAUDE.md 2>/dev/null
If any old version numbers found, fix them.
7b. Agent counts — every file must match actual prompt counts:
ls prompts/tech/ | wc -l # tech count
ls prompts/pm/ | wc -l # pm count
ls prompts/critic/ | wc -l # critic count
Search all docs for stale agent counts (36, 38, or any number that doesn't match the sum above). Fix them.
7c. Command references — every ./command mentioned in docs must exist:
Check each ./command reference against:
SLASH_PREFIXES in src/harness/smart-gateway.tsskills/hooks/
Remove or correct any references to commands that don't exist (e.g. ./team, ./squad).7d. Feature descriptions vs code — spot-check key claims:
TEAM_PRESETS in src/harness/org.tssmart-gateway.tshooks/ and package.jsonALL_ORG_ROLES in org.ts7e. Test count — must match actual:
npm test 2>&1 | tail -3
Update any stale test count references in CLAUDE.md.
Print:
[BW] docs sync complete
README.md: {updated|no changes}
README.ko.md: {updated|no changes}
README.ja.md: {updated|no changes}
docs/plugin-install: {updated|no changes}
docs/skills-guide: {updated|no changes}
CLAUDE.md: {updated|no changes}
old version refs: {count} found and fixed