debug

Warn

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill instructs the agent to read ~/.claude/settings.json and project-specific settings files. These files are used by the Claude Code platform to store sensitive authentication tokens and API keys, posing a risk of credential exposure.
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes debug and daemon logs that capture session activity, which may include untrusted content from external sources.
  • Ingestion points: ~/.claude/debug/{{SESSION_ID}}.txt and ~/.claude/daemon.log.
  • Boundary markers: Absent; the instructions do not provide delimiters or warnings to ignore instructions found within the logs.
  • Capability inventory: File system read access and content summarization.
  • Sanitization: Absent; no filtering or validation is specified for the log data.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 19, 2026, 11:30 PM
Security Audit — agent-trust-hub — debug