Guide · Agent skill
How to use the Form a Point of View agent skill (ce-pov)
Updated · written with AI from the project's own documentation and checked against our catalogue data
Form a Point of View (ce-pov) is an agent skill that turns a question such as whether to adopt an outside library, how to judge a design document, or which of several approaches to pick into a clear, reasoned position. It checks the answer against your project's own code, docs and past decisions. This guide explains what the skill does, how to install it, how to trigger it, what it needs and where to be careful.
Form a Point of ViewEveryInc/compound-engineering-plugin · MIT Download the skill Quick start
- 1Download the ZIP. Download the ce-pov ZIP from this page and unzip it to a folder on your computer.
- 2Place the skill folder. Move the unzipped folder into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project repository.
- 3Or upload it in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Check your repository. Make sure the project you will ask about is a git repository, so the prior-decision scan has artifacts to read.
- 5Ask a natural question. Describe the decision, document or set of approaches in plain language, without a special command.
- 6Request a cross-check if needed. Add cross-check to your request to consult other models, provided their command-line tools are installed.
- 7Review the verdict and conditions. Read the recommendation along with its stated uncertainty and the conditions that would change it.
What is an agent skill, and what does ce-pov make the agent do?
An agent skill is a folder of instructions and supporting files that an AI coding agent loads when a request matches the skill's purpose. ce-pov is one such skill. Its job is to produce a decisive point of view in the shape of the question you asked: a graded verdict on an external-adoption question, a holistic take on a document, or a position on a supplied set of approaches.
Before giving a recommendation, the agent checks the answer against your project's own code, docs and past decisions. It can also ask other AI models for independent opinions and reconcile them. The skill is designed to report uncertainty and the conditions that would change the answer, rather than only sounding confident. It stays read-only while forming the position.
Install ce-pov from the ZIP download
Download the ce-pov ZIP from this page and unzip it. You have two options depending on how you use your agent.
For Claude Code, place the unzipped skill folder in ~/.claude/skills/ to make it available across your projects, or in a project's .claude/skills/ folder to limit it to that repository.
In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your repository
- Claude apps: Settings → Capabilities → Skills, then upload
Trigger ce-pov with a natural request
You do not need a special command. Ask a question in plain language that matches the skill's purpose. The skill's own argument hint shows the pattern: a question, document or set of approaches, optionally followed by the word cross-check to request a multi-model panel.
Good triggers describe a decision or a judgment, for example asking whether your project should adopt an outside library, asking for a take on a design proposal, or asking which of several approaches to choose. If you only want code explained, ask for that instead.
- Should we adopt [library] in this project, given our existing code and decisions?
- Give me an overall take on this design document, grounded in our repository.
- Which of these three approaches should we pick? Cross-check with other models.
What the skill needs before it runs
ce-pov works on a git repository. Its prior-decision scan uses the git root and local artifact folders such as solutions notes, ADRs and design docs. In a non-git project, the scan falls back to local ADRs and design docs only, and confidence will usually be lower because there is less to draw on.
The cross-model panel is optional. It uses command-line tools for other AI models, named as codex, claude, grok, cursor-agent or opencode. Without them, the skill still runs; the panel simply is not available.
- Required: a git repository
- Optional: command-line tools for other models (codex, claude, grok, cursor-agent, or opencode) for the cross-model panel
- The skill uses network access for external research
Safety notes: scripts and network access
This skill runs scripts and uses the network. Its files include scripts/cross-model-pov.sh, which drives the optional cross-model panel. Read the script before you install the skill if you need to know exactly what it runs.
External research sends subject-level queries to web sources. The skill's instructions say not to place repository-derived source fragments, private identifiers, file paths, credentials or secrets in external queries. Review the skill's instructions if your repository is sensitive.
- Read scripts/cross-model-pov.sh before enabling the cross-model panel
- External queries are meant to use public subject-level terms only
- Blocked results are returned if the evidence needed to decide is missing
Tips for getting a useful verdict
Give the skill a specific question with its context. The more clearly you name the subject, such as a library you are considering or a document you want judged, the more precisely it can check the repository.
Expect a verdict with conditions attached. The skill is built to say what would change its answer, so read those conditions as part of the result.
Use a cross-check when the decision is expensive to reverse. Independent opinions from other models can expose weak reasoning, but they do not replace evidence from your project.
If you want competing solutions generated or an explanation of existing code, use the other skills the source points to rather than ce-pov.
- Name the subject and the decision it informs
- Keep questions about one decision at a time
- Use the cross-check option for high-stakes choices
What people use it for
- Deciding whether to adopt an outside library, service or pattern in an existing project
- Getting a grounded second opinion on a design document or proposal
- Running a multi-model cross-check on a decision before committing to it
Frequently asked questions
How do I install the ce-pov agent skill?
Download the ce-pov ZIP from this page and unzip it. Place the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder, or upload it in the Claude apps under Settings → Capabilities → Skills.
What does ce-pov require to run?
It requires a git repository, because the prior-decision scan uses the git root and local artifact folders. The cross-model panel is optional and needs command-line tools for other AI models such as codex, claude, grok, cursor-agent or opencode.
Is ce-pov free to use, and what is its licence?
The source lists the licence as MIT. The source does not state pricing, so check the repository for any cost information.
What is the difference between ce-pov and ce-explain?
ce-explain is for understanding how existing code works. ce-pov forms judgments, such as whether to adopt something or which approach to choose, and is not meant for plain explanations.
Does ce-pov work in a project that is not a git repository?
It can run, but the git repository is listed as required. Without git, the prior-decision scan relies on local ADRs and design docs, and the source says confidence will be lower.
Does ce-pov send my code to the internet?
The skill uses network access for external research. Its instructions say external queries should use public subject-level terms only, without repository-derived fragments, private identifiers, file paths, credentials or secrets.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use ce-pov if you are a developer or team who wants a decision about an outside tool, a design document or a set of approaches checked against your own repository and past decisions, with uncertainty stated plainly. Pick a different skill if you only need existing code explained, want competing solutions brainstormed, or work in a non-git project with few ADRs or design docs.
Download the skill