Skill

performing-paste-site-monitoring-for-credentials

skill:mukul975/anthropic-cybersecurity-skills#performing-paste-site-monitoring-for-credentials

Monitor paste sites like Pastebin and GitHub Gists for leaked credentials, API keys, and sensitive data dumps using automated scraping and keyword matching to detect breaches early.

SkillsMP

Verdict

Suspicious

Risky behavior the analysis could not rule out.

Analyzed Oct 27, 2026, 21:37 UTC at 54a7988

findings

high

highest severity

threat classes

4,532

files scanned

82/83

checks satisfied

package advisories

Why this verdict

suspicious: finding skill:skills/performing-paste-site-monitoring-for-credentials:skills/performing-paste-site-monitoring-for-credentials/scripts/agent.py:68:EXFIL

Security findings 1

Data exfiltration

highContextual riskmedium confidence

bulk_check_domain reads a caller-supplied local employee email-list file and sends each address (up to 50) as a URL path parameter in requests.get calls to the third-party haveibeenpwned.com API, egressing the internal recipient list off-host; generate_exposure_report wraps this same chain.

skills/performing-paste-site-monitoring-for-credentials/scripts/agent.py:68

Path(email_list_file).read_text().splitlines()

vector:code

Counter-argument considered

Attempted refutation fails: the strongest counter is that querying HIBP is benign breach-checking with a reputable hardcoded endpoint and a caller-supplied list, but no carve-out covers third-party egress of an internal email list (the target has no declared claims, and the description covers paste-site scraping/keyword matching, not HIBP API queries), and the source→sink flow is concrete and CLI-reachable, so the contextual-risk finding stands.

Open questions 2

What the analysis could not settle from the code alone.

Frontmatter maps this monitoring skill to T1003/T1555.003 (credentials from password stores/web browsers) and T1110.004 (credential stuffing), techniques no shipped code performs; confirm whether this inflated technique metadata is intended to misstate the skill's coverage.

The declared description covers scraping paste sites with keyword matching, while scripts/agent.py additionally queries haveibeenpwned.com breach/paste APIs; confirm whether HIBP usage is disclosed in the unshown SKILL.md sections before weighting the egress findings.

Data exfiltration

Coverage gaps 2

Parts of the artifact the static analysis could not fully examine.

ledger_finding_without_valid_finding run 1: behavior-chains.0 SRC:1766970c526c35c4

invalid_judge_finding behavior-chains.0