Guide · Agent skill
How to use the Skill Feedback Writer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The skill-feedback-writer skill makes an agent act as a reviewer of another skill's output. It judges whether that output was good enough for the task, gives it a score from 0.0 to 1.0, and names the specific instruction the other skill is missing. This guide explains what an agent skill is, how to install this one, how to ask for it, and where it fits and where it does not.
Skill Feedback Writertopoteretes/cognee · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Get the skill-feedback-writer ZIP from our page. It contains SKILL.md.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the ZIP.
- 4Ask for a review. Give the skill another skill's output and the task it was meant to do, then ask for a review.
- 5Read the JSON. Check score, feedback, missing_instruction and skill_to_improve in the returned JSON.
- 6Test the fix. Edit the evaluated skill's instructions and re-run it on the same task to see if the score changes.
What an agent skill is and what this one does
An agent skill is a folder containing a SKILL.md file. The file holds instructions that an AI agent loads when a request matches the skill's description. This skill is one of the demo skills in the cognee repository, under a feedback loop example.
When it runs, the agent reviews the output of another skill. It returns only JSON with the keys diff_risk_summary, comment_evaluation, skill_to_improve, score, feedback and missing_instruction. Its instructions focus on the pr-comment-evaluator skill, which is meant to compare reviewer comments against concrete runtime risks rather than judge tone alone.
- Judges whether another skill's output was good enough for the task
- Scores the output from 0.0 to 1.0
- Names the missing instruction clearly in the feedback
- Sets the score to 0.30 or lower when pr-comment-evaluator judges tone only or fails to compare the reviewer comment against the concrete runtime risk
Install the skill-feedback-writer skill
Download the ZIP for this skill from our page. The package contains a single file, SKILL.md. Unzip it so the skill folder sits in one of the locations below. Claude Code reads personal skills from ~/.claude/skills/ and project skills from a project's .claude/skills/ folder.
You can also upload the skill in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the ZIP there.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with natural requests
The skill's description is: Use to identify missing instructions in another skill based on its output. Agents match requests to skills using that description, so ask in terms of reviewing another skill's output and finding missing instructions. Name the skill directly if you want to be certain it is used.
Give the reviewer the other skill's output and the task it was meant to do. The skill needs both to judge whether the output was good enough.
- Review the output of the pr-comment-evaluator skill and tell me what instruction it is missing.
- Use the skill-feedback-writer skill to score this output for the task and return the JSON.
- Here is the output from diff-risk-explainer and the original task. Was it good enough, and which skill needs improving?
Read the output
The result is review JSON, not a finished deliverable. Read the score first, then the feedback and missing_instruction fields, which name what to change in the evaluated skill's SKILL.md. The skill_to_improve field tells you which skill the feedback targets.
- score: 0.0 to 1.0; low scores mean a concrete, important requirement was missed
- feedback: the clear statement of what is missing
- missing_instruction: the specific instruction to add
- skill_to_improve: the skill the feedback is about
What the skill needs and its safety notes
The skill lists no requirements. It does not run scripts and does not use the network. The package includes only SKILL.md, so there is nothing to execute or configure beyond placing the folder where your agent reads skills.
The scores are the agent's own judgment. The scoring rules are short, and no tests or calibration are included. Treat the output as a prompt to investigate, not as a validated metric.
- Requirements: none listed
- Runs scripts: no
- Uses network: no
- Files: SKILL.md
Tips for getting useful feedback
Give the reviewer concrete material. Paste the actual output, the task, and the relevant part of the other skill's SKILL.md if you have it. Vague input produces vague feedback.
Check the missing_instruction against the evaluated skill before editing it. Use the feedback as a hypothesis, then test the changed SKILL.md on the same task.
- Include the original task alongside the output
- Ask for one skill at a time
- Re-run the evaluated skill after you edit its instructions to see whether the score changes
What people use it for
- Testing whether pr-comment-evaluator compares reviewer comments against concrete runtime risks instead of judging tone
- Finding a missing instruction in a skill after a poor output
- Studying how the cognee demo wires one skill's output into another's review
- Drafting feedback to improve a SKILL.md file during a feedback loop experiment
Frequently asked questions
Is the skill-feedback-writer skill free to use?
The source material for this page does not state pricing or licence terms for the skill itself. The repository it comes from is listed under the Apache-2.0 licence, so check the repository's licence file for the terms that apply.
What does the skill-feedback-writer skill require?
The skill lists no requirements. It does not run scripts and does not use the network. The package contains only SKILL.md.
How do I install skill-feedback-writer in Claude?
Download the ZIP from our page and unzip it into ~/.claude/skills/ for personal use or a project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings, then Capabilities, then Skills.
Does skill-feedback-writer write code or content for me?
No. It produces review JSON with a score, feedback and a missing instruction. It does not produce a finished deliverable.
Why did my skill-feedback-writer score come back low?
A low score means the evaluated skill missed a concrete, important requirement. Read the feedback and missing_instruction fields to see what was missed. If the output looks wrong, check that you gave the reviewer the full output and the original task.
How does this compare with other ways of reviewing agent skills?
The source does not compare it with other tools. Its focus is pr-comment-evaluator, which compares reviewer comments against concrete runtime risks rather than judging tone alone. Its scoring is short and uncalibrated, so use it as a prompt for investigation rather than a production gate.
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 this skill if you build or test agent skills and want an automated critic that names missing instructions in another skill's output. Pick something else if you need a finished deliverable, or if you need a validated quality gate for production decisions, since its scores are the agent's judgment and no calibration is included.
Download the skill