Skills
skills/eins78/skills/bye/Gen Agent Trust Hub

bye

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute bash commands like ls and sed to locate internal session files and log data within ~/.claude/ for history reconstruction.
  • [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface (Category 8). It ingests untrusted historical conversation data from session JSONL files and logs without sanitization or explicit boundary markers. This data is then used in summaries and git commits. 1. Ingestion points: ~/.claude/projects/*.jsonl and ~/.claude/logs/. 2. Boundary markers: Absent. 3. Capability inventory: File creation, git commit, and git push. 4. Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 08:51 AM