coordinator

Fail

Audited by Socket on Feb 19, 2026

1 alert found:

Malware
MalwareHIGH
SKILL.md

[Skill Scanner] Destructive bash command detected (rm -rf, chmod 777) BENIGN overall in intent: the skill's capabilities (reading manifests, dispatching agents, updating status, optional summary, and cleaning up plan directory) are consistent with a coordinator/orchestrator. However, there is a moderate security risk from the explicit rm -rf .claude/plans/<plan-id>/ instruction and the lack of documented path validation. If an attacker can influence plan-id or working-directory values, they could cause deletion or leakage of files outside the intended plan folder. Recommend validating and constraining paths, requiring explicit confirmation for destructive deletes, and limiting what is included in agent dispatch prompts. LLM verification: Functionally, the coordinator skill fits its purpose of managing modes and orchestrating multi-agent plans. It does not contain obvious obfuscated or directly malicious code in the provided fragment, but it prescribes operational behaviors that are high-risk: automatic destructive deletion (rm -rf) and permissive tool execution without confirmation in TEXT mode. These behaviors can lead to accidental data loss, removal of audit trails, or data exfiltration when combined with remote agents or net

Confidence: 95%Severity: 90%
Audit Metadata
Analyzed At
Feb 19, 2026, 08:50 PM
Package URL
pkg:socket/skills-sh/hjewkes%2Fagent-skills%2Fcoordinator%2F@48034021141c43886cdb8e62fd44a98bafa08f03