skills/aws-samples/ai-skills/ai-plan/Gen Agent Trust Hub

ai-plan

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill contains specific defensive instructions to neutralize prompt injection attempts. It explicitly directs the agent to treat all external content as data and to report any embedded instructions (such as 'ignore previous instructions') as findings rather than executing them.
  • [DATA_EXFILTRATION]: File access is restricted to the current repository, with any operations outside the working directory requiring explicit user approval. Network operations are restricted by the platform's tool configuration, preventing silent data exfiltration.
  • [COMMAND_EXECUTION]: The skill utilizes standard system utilities (date, mkdir, find) for repository management tasks. These are used safely within defined folder structures and are subject to user review via the generated implementation plans.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from feature files (Ingestion points: feature READMEs, research.md) but applies strong mitigations including explicit boundary instructions (Boundary markers: Untrusted Content section). It limits its tools to low-risk operations (Capability inventory: mkdir, date, find) and requires the agent to summarize rather than follow imperatives found in external text (Sanitization: SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 02:43 PM