Reading a skillscan.io report

Every skill page has the same layout. This page explains each part.

Verdict strip

SAFE or ISSUES FOUND, the maximum severity, the source registry, the date of the last scan, and the number of scans. SAFE is the scanner's own threshold decision and a SAFE skill can still have MEDIUM or LOW findings worth reading. The verdict is not a human judgement.

Summary

When the LLM analyzer ran, its overall assessment of the skill is shown as a paragraph. This is the fastest way to understand what the skill does and what concerned the analyzer. Community-requested scans run without the LLM layer and have no summary.

Severity

Level	Meaning
CRITICAL	A pattern that, if real, gives an attacker code execution, credentials, or data. Read the snippet.
HIGH	Serious and likely exploitable in normal use.
MEDIUM	Risky design or disclosure problem; often legitimate functionality with a real cost.
LOW	Hygiene: missing license, unpinned version, loose wording.
INFO	Observations with no direct risk.

Findings

Each finding shows its category (linked to the category page), the rule or analyzer that produced it, the file and line, a snippet, a description, and remediation text. Two fields matter most:

• Analyzer. static findings are signature matches and have the highest false-positive rate. llm and behavioral findings reflect a reading of the whole file. meta fields on an LLM finding show whether a second pass confirmed it.
• Confidence and exploitability (LLM findings only). A finding marked validated with low exploitability is usually a documentation or disclosure issue, not an attack.

What we checked and did not check

Lists the analyzers that ran and the standing blind spots: no execution, no runtime-fetched content, no archive expansion, no image text. See methodology.

Scan history

Every scan of this skill, with finding count and max severity. A change between scans means the skill changed, the scanner changed, or both. The scanner version is recorded on each run so you can tell which.

Links

The registry page, the source repository at the scanned commit, and the raw JSON for this skill. The JSON is the authoritative record; the page is a rendering of it.

Disputing a finding

Open an issue in the repository with the skill name and finding ID. We do not delete findings. Confirmed false positives are listed in the register and annotated on the skill page.