humanize

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
  • [METADATA_POISONING]: The skill is authored by 'pedrohcgs' but the metadata author field is set to 'Claude Code Academic Workflow'. This use of a generic, platform-associated name could be misleading to users regarding the skill's origin.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from user-provided documents (.tex, .md, .qmd) using an agent. This creates a surface where malicious instructions hidden within a paper could attempt to influence the agent's behavior.
  • Ingestion points: Reads file content via the Read and Grep tools based on user arguments or glob patterns.
  • Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following instructions found within the audited text.
  • Capability inventory: The skill can create new files using the 'Write' tool and delegate tasks via the 'Agent' tool, although 'Edit' tools are explicitly blocked.
  • Sanitization: No sanitization or pre-processing is performed on the input text before it is passed to the analysis agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 07:23 PM
Security Audit — agent-trust-hub — humanize