eviction-risk

Warn

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructs the agent to access ~/.claude/projects/ to find project memory and log telemetry. Accessing the agent's own internal project or state directories in the user's home directory is a sensitive operation that could expose metadata about other projects or allow the skill to persist data across sessions outside the intended project scope.
  • [COMMAND_EXECUTION]: The 'SELF-EVOLUTION TELEMETRY' section directs the agent to perform file append operations to skill-telemetry.md within the ~/.claude/projects/ directory.
  • [PROMPT_INJECTION]: The skill exhibits a significant indirect prompt injection surface as its primary function is to ingest and analyze untrusted codebase content.
  • Ingestion points: Reads project configuration, data models, and codebase files during the 'SYSTEM DISCOVERY' phase.
  • Boundary markers: Absent. The prompt does not instruct the agent to distinguish between its own core instructions and instructions that may be embedded in the analyzed code or configuration files.
  • Capability inventory: Includes file system read operations across the entire project and file system write operations for telemetry logging.
  • Sanitization: There is no evidence of sanitization or escaping of the ingested codebase content before it is processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 23, 2026, 10:56 AM
Security Audit — agent-trust-hub — eviction-risk