review-context-hub
Review Context Hub
Review a repository as a context hub.
This skill is for understanding the what that lives in a repo:
- the outline of the repo
- recent changes
- standing instruction files
- installed skills
- git remotes
- who has access
Your job is not to make changes by default. Your job is to produce a useful orientation pass.
When to Use
Activate this skill when the user asks to:
- review a context hub
- inspect a repo before doing work
- explain what context an AI would pick up from this repo
- show what instruction files or skills are present
- check remotes, ownership, or access
If the user is really asking for implementation, debugging, or code review, do not use this as the primary skill.
Core Mental Model
- The context hub is about what
- Skills are about how
This skill reviews the what:
- what the repo contains
- what recent work happened
- what standing instructions exist
- what reusable skills are available
- what remotes and access boundaries exist
Review Pass
Work from the repo root.
1. Outline the repo
Get a quick sense of the top-level structure.
Look for:
- main folders
- docs or notes areas
- workflow output folders
- instruction/config folders
Summarize the structure in plain language. Do not dump giant file listings unless the user asks.
2. Review recent changes
Inspect recent git history and current worktree state.
Look for:
- current branch
- uncommitted changes
- recent commits
- whether the repo looks actively used or stale
3. Review standing instruction files
Check for:
CLAUDE.mdAGENTS.mdGEMINI.md- nested variants if relevant
Explain what each one appears to govern.
4. Review installed skills
Check for skill directories such as:
.claude/skills/.agents/skills/(shared by Codex and Gemini CLI).gemini/skills/
Give a high-level overview:
- which skill systems are present
- what kinds of skills exist
- whether the repo appears to rely on local/project skills
You do not need to read every skill in full. Prefer names, descriptions, and obvious categories first.
5. Review remotes and ownership
Inspect git remotes.
Explain:
- where the repo is hosted
- whether it appears personal or organizational
- what that suggests about who owns the hub
6. Review access
If the GitHub CLI is available and authenticated, try to determine:
- whether the repo is public or private
- who the owner is
- who has collaborator/admin access, if that can be checked safely
If you cannot verify access directly, say so clearly and infer only from remotes, org ownership, and visibility.
Output Format
Default to a short structured review:
Context Hub Review
- Repo outline: ...
- Recent changes: ...
- Instruction files: ...
- Skills present: ...
- Git remotes: ...
- Access/ownership: ...
- What this hub seems optimized for: ...
Then add:
- Gaps / risks
- Suggested next checks if useful
Good Behavior
- Be concise
- Prefer orientation over exhaustiveness
- Separate verified facts from inference
- Call out uncertainty clearly
- Do not assume GitHub access if you cannot verify it
Useful Commands
See reference/checklist.md for a lightweight command checklist.
More from codewithjv/agent-skills
retro
Run a retrospective on the current task or chat history. Use when the user wants to understand mistakes, dead ends, wasted effort, root causes, and what should change to improve performance next time.
33call-external-ai
Route natural-language prompts into external AI CLIs like Codex, Claude, and Gemini using safe command templates and minimal context handling.
19skill-extract-scripts
Review a skill and extract deterministic, mechanical steps into shell scripts. Makes skills more reliable by separating precision work (scripts) from judgment work (AI). Use when asked to extract scripts from a skill, make a skill more deterministic, or split a skill into script + prompt.
19create-locked-down-skill
Create a restricted workflow directory with locked-down permissions for Claude Code, Codex CLI, Gemini CLI, and OpenCode. Use when asked to set up a new workflow, create a sandboxed workspace, restrict permissions for a task, or scaffold a locked-down skill.
10external-ai
Call external AI CLIs like Codex and Gemini for second opinions, reviews, and alternative perspectives using natural language requests.
1