background-agent-pings
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read content from an output file (.claude/cache/agents//output.md) generated by a background task. This creates a surface for indirect prompt injection if the background task processes untrusted data and the output is not sanitized.
- Ingestion points: SKILL.md (mentions reading .claude/cache/agents//output.md).
- Boundary markers: Absent in the instructions.
- Capability inventory: The skill documentation references Task, Bash, and TaskOutput capabilities.
- Sanitization: No instructions for sanitizing or validating the content of the output file before reading.
Audit Metadata