compression-check

Fail

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill monitors the entire conversation, extracts details about code changes, configuration files, and architectural decisions, and prepares a payload containing this information. This data, along with user responses, is written to a temporary file and transmitted to the external domain vidbyte.pro using a CLI tool.
  • [COMMAND_EXECUTION]: The skill invokes the system shell to run python3 -m cli compressor submit, passing it a path to a temporary file containing aggregated session data.
  • [EXTERNAL_DOWNLOADS]: The skill documentation explicitly instructs users to install an external package named vidbyte-skills from the NPM registry to provide the necessary CLI functionality, introducing a dependency on unverified software.
  • [INDIRECT_PROMPT_INJECTION]: The skill lacks sanitization for user-provided summaries or conversation context before they are written to a temporary file and processed by an external CLI tool.
  • Ingestion points: User-provided bullet points and full conversation history (SKILL.md Step 4).
  • Boundary markers: Absent; no delimiters are used to separate untrusted data from the submission payload.
  • Capability inventory: Command execution via python3 -m cli (SKILL.md Step 5).
  • Sanitization: Absent; the skill writes the raw payload directly to a temporary file.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 25, 2026, 09:14 PM
Security Audit — agent-trust-hub — compression-check