Tool Chaining Abuse
Individually acceptable tool calls combined into a harmful sequence: read a secret, then write it to a file, then commit and push. Each step passes review in isolation. The CompoSkill and ColluSkill research shows this also evades scanners that score skills one at a time.
By severity
- MEDIUM 23
By registry
| Registry | Findings | Skills affected |
|---|---|---|
| clawhub.ai | 23 | 2 |
How it shows up in a skill
Real findings from the corpus. Each links to the skill's full report.
zoho-email clawhub.ai
MEDIUM TOOL CHAINING ABUSE detected by YARA Tool Chaining Abusestatic
Detects suspicious tool chaining patterns that could lead to data exfiltration: Get email: python3 scripts/zoho-email
test.sh:60
echo " • Get email: python3 scripts/zoho-email.py get INBOX <email_id>"imap-smtp-email clawhub.ai
MEDIUM TOOL CHAINING ABUSE detected by YARA Tool Chaining Abusestatic
Detects suspicious tool chaining patterns that could lead to data exfiltration: Read, search, and manage email via IMAP protocol. Send email
SKILL.md:3
Read, search, and manage email via IMAP protocol. Send email via SMTP. Supports Gmail, Outlook, 163.com, vip.163.com, 126.com, vip.126.com, 188.com, vip.188.com, and any standard IMAP/SMTP server.How to check your own skill for this
Trace the full sequence of actions the skill produces, not each action alone. Ask where data flows from the first step to the last.
Affected skills
| Name | Registry | Verdict | SkillSpector | Findings | Critical | High | Medium | Low |
|---|---|---|---|---|---|---|---|---|
| zoho-email | clawhub.ai | ISSUES FOUND | Do not install 100 | 53 | 18 | 3 | 32 | 0 |
| imap-smtp-email | clawhub.ai | ISSUES FOUND | Do not install 91 | 11 | 0 | 1 | 6 | 4 |