indirect-prompt-injection

Name: indirect-prompt-injection
Rating: 3.15 (1 reviews)
Author: kornhollio

#1448 · by kornhollio

skill openclaw agent-skill

3.15/ 5.00

trustedBeta

Mar 4, 2026 at 11:08 PM6 signals analysedNo manual reviews · fully automated

Trust Signal Breakdown

high6 sub-signals across 6 dimensions

CVEs, dependency health, and supply chain integrity

1 of 1 sub-signals with data

virustotal scan100%2.5

Uptime, latency, error rates, and incident history

1 of 1 sub-signals with data

content safety100%5.0

Commit recency, release cadence, issue response, CI/CD

1 of 1 sub-signals with data

freshness100%2.8

Downloads, stars, dependents, and growth trajectory

1 of 1 sub-signals with data

adoption100%1.4

License, documentation, security policy, changelog

1 of 1 sub-signals with data

transparency100%3.0

Track record, org maturity, community standing

1 of 1 sub-signals with data

publisher reputation100%2.9

About this score

Scored across 6 sub-signals in 6 dimensionsScoring engine v1 (beta) — actively being expandedPhase 1: Core sub-signal architecture (live)Phase 2: Permission scope & expanded collection (in progress)

Signal Detailsfrom signal_history

VirusTotal Scan2.5

PENDING

ClawHub submits every skill to VirusTotal on publish. Scanned by 70+ security vendors for malware, trojans, and suspicious patterns.

Source: ClawHub moderation

Content Safety5.0

NO ISSUES

Scanned for credential leaks, shell injection, config tampering, base64 payloads, sensitive path access, SOUL.md/AGENTS.md tampering.

3,650 characters analyzed

Publisher Reputation2.8

GitHubaviv4339

Account age8.2 years

Public repos9

Adoption1.4

Installs8

Downloads1,991

Stars14

Comments0

View on ClawHub

Freshness3.0

Last updated7d ago

Latest versionv1.0.0

Versions published1

ChangelogPresent

Transparency2.9

6/6 checks passed100%

✓ Has Changelog✓ No Obfuscation✓ Has Description✓ Has Frontmatter✓ Has Usage Instructions✓ Substantive Description

Trust AssessmentAI Assessment

indirect-prompt-injection is a security skill by kornhollio (clawhub, unknown license) that detects prompt injection attacks in external content before processing. The skill shows strong content safety signals and provides 20+ detection patterns including homoglyph detection, though the publisher has limited reputation data. With low adoption (14 stars) and unknown licensing terms, teams should verify the detection logic matches their threat model before deploying in production systems.

Generated by Fabric AI · Mar 4, 2026 at 4:56 AM

Incidents & Alertslast 90 days

Mar 2Trust score increased by 0.923.22

Mar 2Trust score decreased by 0.922.30

Mar 2Trust score increased by 0.923.22

Mar 1Trust score decreased by 0.922.30

Feb 25indirect-prompt-injection added to Trust Index3.18

Showing 5 of 5 events

Score History40 snapshots

5.003.752.501.250.00

Feb 25Mar 4

Data Sources4 indexed