web-skills-protocol

Low signal level, contained permissions, and limited attack surface.

85/100

1medium

Recommendations

🌐 Outbound network calls — verify the URLs being contacted.

Detected Risks1

Network Callmedium

network_call

Outbound network call

Line 60

Snippets

L60- If the user's intent clearly matches one skill → fetch that SKILL.md

L61- If the intent could match multiple skills → fetch all candidates, pick the best fit

L95

2. **Respect robots.txt.** If `robots.txt` disallows `/skills/` or `/agents/`, do NOT fetch skill files from those paths.

L97

3. **Cache within session.** Fetch `skills.txt`/`agents.txt` once per site per session. Don't re-fetch on every interaction with the same site.

L994. **Don't over-fetch.** Only fetch the SKILL.md files you actually need. Don't download every skill "just in case."

Connectez-vous pour consulter l'analyse detaillee des risques.

$npx agentfend install cmn92mqbj009tu1ip24jl1vsb

Trust Score

88trust

⭐ 3.5K🍴 987

Updated 3 weeks ago

Analyzed

Mar 31, 2026, 3:56 PM

+ 2 previous scans

Compatible with

MCPMCP✓Claude CodeOpenAICodexOCOpenClaw

Skill details

Trust score

88/100

GitHub

Connected

Stars

3.5K

Forks

987

Updated 3 weeks ago

Analyzed Mar 31, 2026, 3:56 PM

Description

Recent scans

Mar 31, 2026, 3:56 PM

Latest analysis

Mar 31, 2026, 3:11 PM

Run 2

Mar 27, 2026, 3:43 PM

Run 1