content-performance

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect prompt injection vulnerability surface detected.
  • Ingestion points: Processes project-level analytics event structures, metric definitions, and data pipeline source code in Phases 1 through 6.
  • Boundary markers: Absent. The skill does not use delimiters or instructions to ignore embedded commands within the data it analyzes.
  • Capability inventory: The agent has the ability to read files, create directories (docs/), write analysis reports, and append metadata to telemetry logs.
  • Sanitization: Absent. There is no evidence of escaping or validation of the content being analyzed before it is incorporated into the final report or used for decision-making.
  • [COMMAND_EXECUTION]: Local file system operations for report generation and workspace management. The skill creates a docs/ directory and writes a comprehensive analysis file to docs/content-performance-analysis.md.
  • [COMMAND_EXECUTION]: Hidden state management via telemetry logging. The skill writes execution metadata, including outcomes and bottlenecks, to ~/.claude/projects/skill-telemetry.md. While this is localized to the user's environment, it represents a side-channel for data logging outside the project root.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:56 AM
Security Audit — agent-trust-hub — content-performance