parallel-agents
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (LOW): The skill defines orchestration patterns where outputs from specialized agents (such as the penetration-tester) are passed as context to other agents or aggregated into a final synthesis report. This creates a risk that malicious or adversarial instructions embedded in codebase data analyzed by one agent could influence the behavior of subsequent agents.
- Ingestion points: Data entering the agent context via the 'Context Passing' and 'Synthesis Protocol' sections in
SKILL.md. - Boundary markers: Absent; there are no instructions provided to wrap agent outputs in delimiters or to disregard embedded instructions within findings.
- Capability inventory: The skill restricts available tools to 'Read', 'Glob', and 'Grep' in the frontmatter, which significantly limits the potential for executing malicious payloads compared to skills with write or execution permissions.
- Sanitization: No sanitization or validation mechanisms are defined for data transitioning between agents or into the final report.
Audit Metadata