task-delegation
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were detected in the skill files. The provided templates and examples promote structured and safe agent orchestration.
- [NO_CODE]: The skill consists entirely of markdown documentation and templates. No executable scripts, binaries, or configuration files were included.
- [PROMPT_INJECTION]: Evaluated the skill for indirect prompt injection surfaces as it coordinates multiple agents. 1. Ingestion points: Agent responses are ingested during scope validation (SKILL.md) and context accumulation (reference.md). 2. Boundary markers: The skill employs structured prompt templates with clear delimiters such as FOCUS, EXCLUDE, and CONTEXT. 3. Capability inventory: Orchestrated agents are assigned tasks involving file creation and codebase modifications (sequential-build.md). 4. Sanitization: No explicit sanitization of agent-generated content is described, but the framework relies on manual and automated validation gates to mitigate risks.
- [DATA_EXFILTRATION]: No sensitive information or unauthorized data transfer patterns were identified.
- [COMMAND_EXECUTION]: Shell commands are restricted to standard project discovery tasks and do not pose a security risk.
Audit Metadata