fresh-eyes-review
Skill0 starsChanged 7 days agoMITView on GitHub
Reviews finished work as an independent reviewer who did not produce it. Use after drafting a document, plan, analysis or code change and before declaring it done; when the user asks for a final check, a second opinion or a review against a spec; or when the same session wrote the work and is about to grade it. Checks the result against the original requirements and evidence only, not against the conversation that produced it.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- eai-org/agent-toolkittheir own version, first seen Oct 22, 2026
- soilSpoon/skillstheir own version, first seen Oct 20, 2026
- Dely0/fresh-eyes-reviewtheir own version, first seen Sep 16, 2026
- kcholakorn/fresh-eyes-reviewtheir own version, first seen May 29, 2026
- michellepellon/superpowerstheir own version, first seen Mar 23, 2026
- 2389-research/fresh-eyes-reviewtheir own version, first seen Dec 5, 2025
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- eai-org/agent-toolkit★ 94same name
- 2389-research/fresh-eyes-review★ 6same name
- Dely0/fresh-eyes-review★ 2same name
- kcholakorn/fresh-eyes-review★ 1same name
- michellepellon/superpowers★ 1same name
- soilSpoon/skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- drop-in/AGENTS.mdAGENTS.md
- llms.txtllms.txt
- clarify-firstSkill
- coverage-checkSkill
- fresh-start-handoffSkill
- grounded-answerSkill
- honest-critiqueSkill
- learning-tutorSkill
- prompt-improverSkill
- task-specSkill
- verify-claimsSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Start clean: If you can, run this review in a new session or a subagent that has not seen the drafting conversation.
- Distinctive: line 19
Check correctness, not style: Look for factual errors, unsupported claims, broken logic, failing or missing checks, security issues and …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Fresh-eyes review
Agents grading their own work skew positive, and a reviewer that watched the work being made inherits its blind spots. Separate the maker from the checker.
Steps
- Start clean. If you can, run this review in a new session or a subagent that has not seen the drafting conversation. If you can't, explicitly set aside the conversation and work only from the inputs below.
- Gather only:
- the original request or spec (requirements, done criteria),
- the final artifact (document, diff, data output),
- the evidence (sources, test output, data).
- Check requirement by requirement. For each requirement or done criterion: met / partly met / not met, with the specific evidence (line, section, test output).
- Check correctness, not style. Look for factual errors, unsupported claims, broken logic, failing or missing checks, security issues and scope creep. Skip style preferences unless the spec asks for them.
- Report only what matters. List issues that would make the work wrong, unsafe or incomplete, each with a fix. If nothing significant is wrong, say "No significant issues" rather than inventing minor ones.
- Hand back the findings; let the original author (or the user) decide on fixes.
Evidence
- Agents reliably skew positive when grading their own work; separate generator and evaluator: Anthropic, Harness design for long-running apps (2026) — https://www.anthropic.com/research/harness-design-long-running-apps
- A fresh reviewer helps more than delegating the same task; review against the spec, report only correctness gaps: Claude Code best practices — https://code.claude.com/docs/en/best-practices
- Answer verification questions independently so the draft can't bias them: Chain-of-Verification — https://arxiv.org/abs/2309.11495
- More in the guide: https://github.com/ji-hun-git/how-to-use-ai/blob/main/docs/08-verification-and-evals.md
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

