claude-devfleet

Pass

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a multi-agent orchestration workflow that is inherently vulnerable to indirect prompt injection due to its handling of untrusted task descriptions.
  • Ingestion points: The skill ingests untrusted instructions through the plan_project(prompt) and create_mission(prompt) tools defined in SKILL.md.
  • Boundary markers: There are no instructions or delimiters defined to separate user-provided mission prompts from the agents' internal system instructions, which could allow task descriptions to override agent behavior.
  • Capability inventory: Dispatched agents are described in SKILL.md as having 'full tooling' and operating within git worktrees, providing a significant capability surface for any successfully injected instructions.
  • Sanitization: No sanitization, validation, or escaping mechanisms are specified for the prompts passed between the orchestrator and the dispatched agents.
  • [NO_CODE]: The skill consists entirely of markdown instructions and configuration metadata. It contains no scripts, binaries, or directly executable code files, which limits the risk of direct malicious execution from the skill itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 14, 2026, 02:57 PM