菜单
Use when onboarding new Codex features, reviewing external best-practice repos, or expanding the daily learning catalog. Covers: gap analysis against native CC capabilities, tips catalog expansion, daily-learning script, native .Codex/ structure (agents, commands, settings).
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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/vamseeachanta/workspace-hub --skill cc-best-practice-adoption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。
基于 SOC 职业分类
正在显示 SKILL.md
| name | cc-best-practice-adoption |
| description | Use when onboarding new Codex features, reviewing external best-practice repos, or expanding the daily learning catalog. Covers: gap analysis against native CC capabilities, tips catalog expansion, daily-learning script, native .Codex/ structure (agents, commands, settings). |
| version | 1.0.0 |
| updated | "2026-04-03T00:00:00.000Z" |
| category | workspace-hub |
| type | skill |
| tags | ["learning","onboarding","Codex","best-practice","daily-cadence"] |
| related_skills | ["comprehensive-learning"] |
| platforms | ["linux"] |
Workflow for analyzing external Codex best-practice sources, distilling actionable gaps in our ecosystem, and wiring daily learning into the session lifecycle.
Critical lesson learned: /powerup, /insights, /stats, /context, /simplify,
/batch, /debug, /diff, /security-review are all native CC commands (v2.1.90+).
We wasted effort planning to build custom versions before discovering they exist.
Rule: Before creating any slash command, run /help or check
Codex-best-practice/best-practice/Codex-commands.md
for the 64 native commands.
Our ecosystem has 691 Hermes-style skills but ZERO native CC format. They are different:
.Codex/skills/<name>/SKILL.md with CC frontmatter (context: fork, paths, model, effort, hooks)~/.hermes/skills/<name>/SKILL.md with Hermes frontmatter/skills menu and support auto-discovery.config/workflow-tips/tips-catalog.yaml uses a minimal structure parsed by
daily-learning.py without PyYAML. Don't add complex YAML features (anchors, flow
sequences in values). Keep entries as simple key: value pairs.
daily-learning.py uses md5(YYYY-MM-DD) as seed so the same 2 tips show all day.
Running it multiple times won't cycle tips — that's intentional (spaced repetition).
cd /tmp && git clone --depth 1 <repo-url>
Key files to read first:
Compare external catalog against:
config/workflow-tips/tips-catalog.yaml — what tips are we missing?.Codex/settings.json — what settings are we not using?.Codex/agents/ — do we have native agent definitions?.Codex/commands/ — do we have native command definitions?Output a structured list: ALREADY DONE / STILL MISSING / NEW TO ADD.
Append new tips to config/workflow-tips/tips-catalog.yaml following the format:
- id: <prefix>-<slug> # cc- | eco- | gsd- | bp-
category: <category> # Codex | ecosystem | gsd | practice
name: <Short Name>
oneliner: "<one sentence>"
try_it: "<copy-pasteable command>"
source: "<provenance>"
tags: [tag1, tag2]
added: YYYY-MM-DD
Prefixes: cc- (Codex native), eco- (ecosystem), gsd- (GSD), bp- (best practice).
If new tip categories are added, update scripts/productivity/daily-learning.py:
show_daily_tips()generate_practice() for high-value tipsCreate files using native CC frontmatter (not Hermes format):
Agents (.Codex/agents/<name>.md):
---
name: <name>
description: "<when to invoke — use PROACTIVELY for auto-invocation>"
model: haiku|sonnet|opus
tools: Read, Write, Edit, Bash, Glob, Grep
color: cyan|red|green|blue|magenta
memory: project
effort: low|medium|high|max
isolation: worktree # optional
skills: # optional — preloaded into context
- skill-name
---
Commands (.Codex/commands/<name>.md):
---
name: <name>
description: "<what it does>"
model: haiku|sonnet|opus
effort: low|medium|high
allowed-tools: Read, Bash, Agent(explorer)
---
Use !`command` for dynamic shell injection — output injected into prompt.
Settings (.Codex/settings.json) — key fields often missed:
outputStyle: "Explanatory" — insight boxeseffortLevel: "high" — default reasoning depthworktree.symlinkDirectories — fast worktree startupspinnerVerbs — domain-relevant waiting messagesspinnerTipsOverride — ecosystem tips while spinningAdd a SessionStart hook for automatic tip surfacing:
{
"type": "command",
"command": "bash .Codex/hooks/daily-learning-tip.sh 2>/dev/null || true",
"timeout": 3,
"statusMessage": "Daily learning tip"
}
uv run scripts/productivity/daily-learning.py --progress # category coverage
uv run scripts/productivity/daily-learning.py --categories # browse all tips
| File | Purpose |
|---|---|
config/workflow-tips/tips-catalog.yaml | All tips (67+), 4 categories |
config/workflow-tips/tip-history.yaml | Shown-tip tracking, 30-day window |
scripts/productivity/daily-learning.py | Daily tip picker + practice exercises |
.Codex/hooks/daily-learning-tip.sh | SessionStart hook, 1 tip on start |
.Codex/agents/*.md | Native CC agent definitions |
.Codex/commands/*.md | Native CC slash commands |
.Codex/settings.json | CC configuration |
PreToolUse hook entry in .Codex/settings.json
for Write|Edit|MultiEdit, then delegate the decision to a reusable shell checker
(example: .Codex/hooks/config-protection-pretooluse.sh -> scripts/enforcement/check-config-protection.sh).
This keeps policy logic testable outside the hook wrapper.pyproject.toml
and other low-risk config touches, but block broad weakening patterns like ignore,
extend-ignore, per-file-ignores, disable, skip, off, and block removal of
core safety-gate text from AGENTS.md / AGENTS.md unless an explicit env bypass is set.CONFIG_PROTECTION_APPROVED=1).