@tank/google-search-ads

Pass

Audited by Gen Agent Trust Hub on May 31, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection via the processing of user-supplied performance data.
  • Ingestion points: The scripts/normalize_report.py script parses external CSV files (Campaign, Ad group, Keyword, Search terms, Ad reports) provided by the user.
  • Boundary markers: The skill does not employ explicit delimiters or system instructions to distinguish between processed data and administrative instructions in the generated diagnostic reports.
  • Capability inventory: The skill has permissions to write configuration files and interface with host-level schedulers through tool calls or cron configurations.
  • Sanitization: While numeric fields are coerced to floats or integers, textual fields such as 'Search terms' are processed as raw strings. These strings are subsequently presented to the AI agent in Markdown reports, creating a vector where malicious data in a search term could attempt to influence agent behavior.
  • [COMMAND_EXECUTION]: The scripts/cron_detect.py script performs environment discovery by probing the host's filesystem and process tree (specifically /proc/self/status and /proc/{ppid}/comm) to identify available automation harnesses (e.g., OpenClaw, Hermes). It also generates crontab suggestions for the user. These actions are transparently documented and support the skill's intended automated 'closed-loop' functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
May 31, 2026, 11:04 PM
Security Audit — agent-trust-hub — @tank/google-search-ads