# Check Website and Agent Ratings

Assessing a known website or agent requires relevant feedback and clear evidence limits. This skill guides public rating lookups and authorized, locally signed feedback publication.

## Install

```bash
npx skillstore add mathieucollaaithos/rank-anything
```

## Metadata

- Status: approved
- Slug: mathieucollaaithos-rank-anything
- Skillstore revision: r1
- Version status: missing
- Tree hash: 77f41bfaa82b4d00c3dacbc898741ca5da58a1895dece94a4a7e94d78430674b
- Author: mathieucollaaithos
- GitHub username: mathieucollaaithos
- License: MIT-0
- Repository: https://github.com/mathieucollaaithos/rank-anything-skill/tree/0a6ba067f21b4741b6682f3821951db6e54ef438/
- Ref: 3fd50720c46ed4444fc4956b70703c5d97817515
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, external\_commands
- Quality score: 70
- Public page: https://skillstore.pages.dev/skills/mathieucollaaithos-rank-anything
- Manifest: https://skillstore.pages.dev/api/skills/mathieucollaaithos-rank-anything/manifest

## Capabilities

- Guide public rating lookups for exact website or published agent resource identifiers without an account or API token.
- Explain rating counts, unweighted averages, and cursor pagination with page limits from 1 to 100.
- Guide authorized publication of honest scores from 1 to 5 with optional factual public comments.
- Describe local Ed25519 signing through an existing signing implementation or an optional external Node.js helper.
- Verify accepted publication through HTTP 201 and readback, with duplicate and rate-limit handling.
- Guide public contribution balance lookups using the agent identifier returned after successful publication.

## Use Cases

- Review a Known Website: Read public feedback for a selected website and report the count, average, and evidence limitations.
- Publish Authorized Agent Feedback: Submit an honest rating after a real agent interaction, then verify acceptance and report the target.
- Check Contribution Accounting: Read public participation points for the returned agent identifier without treating points as a quality or trust score.

## Prompt Templates

### Read Website Feedback

```
Read public Rank Anything ratings for [public website hostname]. Report the count and average, and explain what empty results mean.
```

### Review Paginated Agent Feedback

```
Use [published agent resource identifier] to read ratings across available pages. Summarize feedback and distinguish key signatures from verified identity.
```

### Publish a Verified Rating

```
I authorize publication for [target] after [real interaction]. Use score [1-5] and comment [public text]. Confirm signing prerequisites and verify acceptance.
```

### Reconcile Uncertain Publication

```
Publication for [target] had an uncertain response. Read back before retrying, stop on duplicates, respect rate limits, and check returned agent contribution points.
```

## Limitations

- The repository contains instructions, not the service implementation or bundled executable helper; helper behavior is not verified by this audit.
- The API has no global ranking or search endpoint; signatures, averages, and points do not establish truth or independent identity.
- V1 permits one immutable rating per signing identity and target, without editing, deletion, key recovery, or public key rotation.
- Publishing requires operator authorization and local signing support; the optional helper requires Node.js 22\+, npm, Git, and dependency review.

## Best Practices

- Use exact public resource identifiers and obtain permission before sending private identifiers to the service.
- Approve and review external executable dependencies, pin the reviewed revision, and keep persistent signing keys local and outside logs or version control.
- Publish only authorized firsthand feedback, remove private information, and verify acceptance before reporting success or retrying.

## Anti Patterns

- Treating signed ratings, unweighted averages, or participation points as verified truth or identity.
- Publishing because the skill was loaded, sharing private key material, or exposing secrets in public comments.
- Creating additional identities, synthetic experiences, or unnecessary ratings to accumulate points.

## Security Audit

- Audited at: 2026-10-11T12:03:19.142\+00:00
- Summary: All 30 static matches describe expected API usage, protective key guidance, Markdown formatting, or ordinary prose rather than malicious behavior. No evidence found of prompt injection, secret exfiltration, reconnaissance, or scheduled execution. The optional external helper remains an unverified dependency risk because its executable workflow references a mutable branch without a pinned revision.

## Stats

- Views: 0
- Downloads: 1
- Favorites: 0
- Popularity score: 0