pensar-attack-surface

Fail

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the user to install its core tool using the command curl -fsSL https://pensarai.com/install.sh | bash, which involves downloading and executing a script directly from the author's domain.
  • [COMMAND_EXECUTION]: The skill relies on the pensar CLI to perform actions such as authenticating sessions, modifying application structures, and dispatching penetration tests.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by allowing users to define security objectives, business logic descriptions, and threat models that are intended to be consumed by other automated agents.
  • Ingestion points: Metadata fields and command arguments such as --objective, --business-logic, and --threat-model described in SKILL.md.
  • Boundary markers: None identified; the skill does not specify the use of delimiters or 'ignore' instructions for these fields.
  • Capability inventory: The skill uses the pensar CLI to manipulate workspace data and trigger pentest runs.
  • Sanitization: No sanitization or validation logic is specified for the interpolated user strings.
Recommendations
  • HIGH: Downloads and executes remote code from: https://pensarai.com/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 4, 2026, 04:25 AM
Security Audit — agent-trust-hub — pensar-attack-surface