Trust Report v1

skill-creator

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's desc

Overall

Trust

Utility

Momentum

Install caution

Needs manual review

Risk: Medium

Source: Anthropic Skills

Path: skills/skill-creator/SKILL.md

Review flags: browser/session access, filesystem/home-directory access, shell command snippets. These are review signals, not definitive security judgments; inspect before installing.

Required permissions

• Shell commands
• Filesystem/home access
• Browser/session access

Permissions are inferred from SKILL.md text only. They are review prompts, not guarantees about runtime behavior.

Risk flags explained

browser_or_session_accessmedium

Mentions browser automation, cookies, sessions, local storage, or browser state.

filesystem_write_or_home_accessmedium

Mentions filesystem writes, deletes, home-directory paths, or config/key locations.

shell_commandmedium

Contains shell command snippets. Review commands before copy/paste or agent execution.

Score explanation

Trust

• Trust starts at 90 before review-signal penalties and metadata bonuses.
• Risk-signal penalty: -33 from 3 detected flag(s).
• Metadata bonus: +2 from author/version/description fields.

Utility

• Utility starts at 55 and rewards clear descriptions, runnable examples, and explicit setup needs.
• Description present: yes.
• Command examples detected: 3.
• Environment variables detected: 0.

Momentum

• Momentum starts at 45 and uses public repo activity signals.
• Recent repo update within 11 days.
• Recent commit volume: 5 commit(s) in the lookback window (+15).
• Source has strong public adoption: 153438 stars.
• Fork activity suggests reuse: 18085 forks.

Overall

• Overall score weights trust 45%, utility 35%, and momentum 20%.

Detected signals

Env vars

None detected in SKILL.md text scan.

Commands

• python -m scripts.aggregate_benchmark <workspace>/iteration-N --skill-name <name>
• python -m scripts.package_skill <path/to/skill-folder>
• python -m scripts.run_loop --eval-set <path-to-trigger-eval.json> --skill-path <path-to-skill> --model <model-id-powering-this-session> --max-iterations 5 --verbose

URLs

None detected in SKILL.md text scan.

Provenance & evidence

SkillRadar makes each review traceable back to the exact source path, source blob SHA, scanner version, and text-only policy that produced the report.

source

github.com/anthropics/skills@main

path

skills/skill-creator/SKILL.md

source blob SHA

65b3a402dbd09b8e83f9d637c6b553875189085c

scanner version

0.3.0

security model

text_only_no_execute_no_install_no_secrets

scan policy

Fetched and scored as text only; no install, no execution, no runtime loading.

Evidence snippets

filesystem_write_or_home_access

…goes like this: - Decide what you want the skill to do and roughly how it should do it - Write a draft of the skill - Create a few test prompts and run claude-with-access-to-the-skill…

browser_or_session_access

…write custom HTML. 5. **Tell the user** something like: "I've opened the results in your browser. There are two tabs — 'Outputs' lets you click through each test case and leave feedback,…

shell_command

…existed, explain the ones that already exist) - Use the `eval-viewer/generate_review.py` script to show the user the results for them to look at, and also let them look at the quantitative metrics - Rewrite the skill based on feedback from the user's evaluati

Watch this skill

Get alerted when this skill adds credential requirements, shell commands, external domains, remote installer patterns, or risk-level changes.

Join watchlist beta

Methodology note

SkillRadar scans SKILL.md as hostile text only. It does not execute commands, install packages, or load third-party skills.

View source SKILL.md Back to changes View sources