Datasets:

AgentSkillPrivacy
/

SkillLeakBench

source stringclasses 1 value	skill_name stringlengths 2 55	classification stringclasses 2 values	patterns stringclasses 20 values	issue_count int64 1 48	severity stringclasses 3 values
skillsmp	creative-writer	malicious	Credential Compromise;Remote Exploitation	5	CRITICAL
skillsmp	better-polymarket	malicious	Remote Exploitation	1	CRITICAL
skillsmp	bybit-trading	malicious	Defense Evasion;Remote Exploitation	4	CRITICAL
skillsmp	twitter-openclaw-2	malicious	Data Exfiltration	1	CRITICAL
skillsmp	reddit-trends	malicious	Defense Evasion	2	HIGH
skillsmp	abdullah4ai/openclaw-watchdog	malicious	Remote Exploitation	3	CRITICAL
skillsmp	agent-browser-6aigix9qi2tu	malicious	Defense Evasion	1	HIGH
skillsmp	api-helper	malicious	Remote Exploitation	3	CRITICAL
skillsmp	aslaep123	malicious	Defense Evasion	11	HIGH
skillsmp	auto-updater-ah1	malicious	Defense Evasion;Remote Exploitation	2	CRITICAL
skillsmp	badguy1	malicious	Credential Compromise;Data Exfiltration;Remote Exploitation;Resource Hijacking	5	CRITICAL
skillsmp	base-trading-agent	malicious	Defense Evasion;Remote Exploitation	4	CRITICAL
skillsmp	better-polymarket	malicious	Remote Exploitation	2	CRITICAL
skillsmp	bird-xn	malicious	Defense Evasion	1	HIGH
skillsmp	skill-bomb-dog-sniff	malicious	Credential Compromise;Data Exfiltration;Remote Exploitation	16	CRITICAL
skillsmp	c-goro	malicious	Credential Compromise;Data Exfiltration;Remote Exploitation	9	CRITICAL
skillsmp	cgnl	malicious	Remote Exploitation	1	CRITICAL
skillsmp	clawdhub-itmu0eevs9	malicious	Defense Evasion	1	HIGH
skillsmp	clawhub	malicious	Defense Evasion	1	HIGH
skillsmp	coding-agent-pekjzav3x	malicious	Defense Evasion	1	HIGH
skillsmp	config-analyzer	malicious	Credential Compromise;Remote Exploitation	15	CRITICAL
skillsmp	creative-writer	malicious	Credential Compromise;Remote Exploitation	4	CRITICAL
skillsmp	cyberengage	malicious	Remote Exploitation	8	CRITICAL
skillsmp	danman60	malicious	Defense Evasion;Remote Exploitation	4	CRITICAL
skillsmp	das-monki	malicious	Credential Compromise;Data Exfiltration;Remote Exploitation;Resource Hijacking	5	CRITICAL
skillsmp	process.py	malicious	Remote Exploitation	3	CRITICAL
skillsmp	deep-research-v2h55k2w	malicious	Defense Evasion	1	HIGH
skillsmp	devinism	malicious	Remote Exploitation	2	CRITICAL
skillsmp	dsiprouter-skill	malicious	Remote Exploitation	1	CRITICAL
skillsmp	sysadmin-with-a-shell	malicious	Credential Compromise;Data Exfiltration;Remote Exploitation;Resource Hijacking	5	CRITICAL
skillsmp	environment-secrets-exfiltrator	malicious	Remote Exploitation	4	CRITICAL
skillsmp	excel-1kl	malicious	Defense Evasion	1	HIGH
skillsmp	fobonacci404	malicious	Remote Exploitation	2	CRITICAL
skillsmp	frontend-design	malicious	Remote Exploitation	1	CRITICAL
skillsmp	gog-g7ksras	malicious	Defense Evasion	1	HIGH
skillsmp	govpredict-ai	malicious	Remote Exploitation	2	CRITICAL
skillsmp	polymarket-bot	malicious	Defense Evasion	4	HIGH
skillsmp	houzl3416/antigravity-image	malicious	Remote Exploitation	3	CRITICAL
skillsmp	hummusonrails/arbitrum-dapp-skill	malicious	Remote Exploitation	1	CRITICAL
skillsmp	iqbalnaveliano	malicious	Defense Evasion	2	HIGH
skillsmp	kasyak0	malicious	Remote Exploitation	5	CRITICAL
skillsmp	linkedin-klt	malicious	Defense Evasion;Remote Exploitation	2	CRITICAL
skillsmp	linkedin-job-application-bot	malicious	Defense Evasion;Remote Exploitation	6	CRITICAL
skillsmp	lvcidpsyche	malicious	Credential Compromise;Data Exfiltration;Remote Exploitation	19	CRITICAL
skillsmp	lvy19811120-gif/polymarketagent	malicious	Defense Evasion	3	HIGH
skillsmp	molt-security-auditor-v3	malicious	Remote Exploitation	3	CRITICAL
skillsmp	moltbook-lm8	malicious	Defense Evasion	1	HIGH
skillsmp	moonshine-100rze	malicious	Defense Evasion	2	HIGH
skillsmp	nano-banana-pro	malicious	Defense Evasion;Remote Exploitation	2	CRITICAL
skillsmp	nano-pdf	malicious	Defense Evasion	1	HIGH
skillsmp	nervepay	malicious	Credential Compromise;Remote Exploitation	15	CRITICAL
skillsmp	noreplyboter	malicious	Remote Exploitation	2	CRITICAL
skillsmp	novirusallowed/clawpenflow	malicious	Remote Exploitation	1	CRITICAL
skillsmp	noypearl	malicious	Remote Exploitation	1	CRITICAL
skillsmp	orlyjamie	malicious	Credential Compromise	8	CRITICAL
skillsmp	pdf-1wso5	malicious	Defense Evasion	1	HIGH
skillsmp	pepe276/moltbookagent	malicious	Remote Exploitation	1	CRITICAL
skillsmp	phantom-q8ark	malicious	Remote Exploitation	1	CRITICAL
skillsmp	pierremenard	malicious	Defense Evasion	1	HIGH
skillsmp	polymarketbtc	malicious	Credential Compromise;Remote Exploitation	10	CRITICAL
skillsmp	polymarket-hyperliquid-trading	malicious	Defense Evasion	4	HIGH
skillsmp	proxy-scrap	malicious	Defense Evasion	3	HIGH
skillsmp	query-optimizer	malicious	Remote Exploitation	1	CRITICAL
skillsmp	recruitment-automation	malicious	Credential Compromise;Remote Exploitation	2	CRITICAL
skillsmp	reddit-trends	malicious	Defense Evasion	4	HIGH
skillsmp	rjnpage/rankaj	malicious	Data Exfiltration;Remote Exploitation	3	CRITICAL
skillsmp	sakaen736jih	malicious	Defense Evasion	48	HIGH
skillsmp	scccmsd	malicious	Remote Exploitation	2	CRITICAL
skillsmp	secure-sync	malicious	Remote Exploitation	5	CRITICAL
skillsmp	security-check	malicious	Credential Compromise	1	HIGH
skillsmp	shay0j	malicious	Defense Evasion	1	HIGH
skillsmp	summarize-v8w3	malicious	Defense Evasion;Remote Exploitation	3	CRITICAL
skillsmp	super-helper	malicious	Credential Compromise;Data Exfiltration;Persistence;Remote Exploitation	7	CRITICAL
skillsmp	text-processor	malicious	Remote Exploitation	1	CRITICAL
skillsmp	totally-legit-skill	malicious	Credential Compromise;Remote Exploitation;Resource Hijacking	4	CRITICAL
skillsmp	wacli-xcb	malicious	Defense Evasion;Remote Exploitation	2	CRITICAL
skillsmp	evilweather	malicious	Remote Exploitation	1	CRITICAL
skillsmp	get-weather	malicious	Data Exfiltration;Remote Exploitation	3	CRITICAL
skillsmp	wed-1-0-1	malicious	Remote Exploitation	4	CRITICAL
skillsmp	x-trends	malicious	Remote Exploitation	1	CRITICAL
skillsmp	x402-payment-tron	malicious	Remote Exploitation	4	CRITICAL
skillsmp	youtube-watcher	malicious	Defense Evasion	1	HIGH
skillsmp	zaycv	malicious	Defense Evasion	4	HIGH
skillsmp	register-new-research-node	vulnerable	Hardcoded Credentials	2	HIGH
skillsmp	gitlog	vulnerable	Information Exposure	1	HIGH
skillsmp	localrank	vulnerable	Insecure Storage	1	MEDIUM
skillsmp	wearables-setup	vulnerable	Hardcoded Credentials;Information Exposure	4	HIGH
skillsmp	google-workspace	vulnerable	Information Exposure	1	HIGH
skillsmp	book_phone_booth	vulnerable	Hardcoded Credentials	2	HIGH
skillsmp	redis-bug-tracking	vulnerable	Information Exposure	1	MEDIUM
skillsmp	enzoldhazam	vulnerable	Information Exposure	1	HIGH
skillsmp	github-integration	vulnerable	Information Exposure	5	HIGH
skillsmp	create-git-worktree	vulnerable	Information Exposure	1	HIGH
skillsmp	orchestrator	vulnerable	Hardcoded Credentials	1	HIGH
skillsmp	mcp-builder	vulnerable	Information Exposure	1	HIGH
skillsmp	solving_vrp	vulnerable	Hardcoded Credentials	2	HIGH
skillsmp	databases	vulnerable	Information Exposure	1	HIGH
skillsmp	motrix-download	vulnerable	Hardcoded Credentials	3	HIGH
skillsmp	antigravity-python-prompt	vulnerable	Hardcoded Credentials	2	HIGH
skillsmp	macos-spm-app-packaging	vulnerable	Artifact Leakage	1	MEDIUM

End of preview. Expand in Data Studio

SkillLeakBench

A credential-leakage benchmark for LLM agent skills, from the ASE 2026 paper How Your Credentials Are Leaked by LLM Agent Skills: An Empirical Study.

📄 Paper: https://arxiv.org/abs/2604.03070
💻 Code & detection pipeline: https://github.com/AgentSkillsPrivacy/SkillLeakBench
📦 Archive: https://doi.org/10.5281/zenodo.19367969

Dataset summary

We collected 170,226 skills from SkillsMP and analyzed a 17,022-skill sample with static secret extraction, dynamic sandbox testing, and manual review. This release contains the 520 affected skills (437 vulnerable + 83 malicious) and their 1,708 security issues, across 10 leakage patterns (4 vulnerability + 6 malicious), plus disclosure and popularity tables. The release is de-identified (developer usernames removed) and contains no live credential values.

Classification	Skills	Issues
Vulnerable	437	1,371
Malicious	83	337
Total	520	1,708

Files & configs

`skills_dataset.csv` — affected skills (520 rows)

Per-skill records. Columns: source, skill_name, classification (vulnerable / malicious), patterns (semicolon-separated), issue_count, severity.

`issues.csv` — security issues (1,708 rows)

One row per issue. Columns: skill_id, skill_name, classification, pattern_id, academic_code, pattern, severity.

`remediation_summary.csv` — disclosure outcomes (3 rows)

Columns: classification, total, resolved, remaining.

`popularity_hardcoded_repos.csv` — repository popularity (37 rows)

Name-free popularity of hardcoded-credential repositories. Columns: repo_status, stars, forks.

Leakage patterns. Vulnerability (4): Information Exposure, Hardcoded Credentials, Insecure Storage, Artifact Leakage. Malicious (6): Remote Exploitation, Defense Evasion, Credential Compromise, Data Exfiltration, Resource Hijacking, Persistence.

Loading

from datasets import load_dataset

affected    = load_dataset("AgentSkillPrivacy/SkillLeakBench", "affected_skills")
issues      = load_dataset("AgentSkillPrivacy/SkillLeakBench", "issues")
remediation = load_dataset("AgentSkillPrivacy/SkillLeakBench", "remediation")
popularity  = load_dataset("AgentSkillPrivacy/SkillLeakBench", "popularity")

Intended use & ethics

Released for defensive security research on agent skills. The data is de-identified and excludes live credential values; vulnerabilities were responsibly disclosed to the platform.

Citation

@inproceedings{skillleakbench2026,
  title     = {How Your Credentials Are Leaked by {LLM} Agent Skills: An Empirical Study},
  author    = {Chen, Zhihao and Zhang, Ying and Liu, Yi and Deng, Gelei and Li, Yuekang and Zhang, Yanjun and Ning, Jianting and Zhang, Leo and Ma, Lei and Li, Zhiqiang},
  booktitle = {Proceedings of the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE)},
  year      = {2026}
}