systematic-debugging

Pass

Audited by Gen Agent Trust Hub on Oct 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides a standalone Bash script find-polluter.sh which executes npm test and find commands to locate state-polluting test files. Additionally, SKILL.md suggests using system tools like codesign, security, and env for gathering evidence during the debugging process.\n- [DATA_EXFILTRATION]: Instructions in SKILL.md encourage the inspection of sensitive system data for diagnostic purposes, including environment variables (env | grep IDENTITY) and macOS keychain identities (security list-keychains). While intended for local debugging, this creates a potential data exposure risk if the agent includes these values in its output logs or responses.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data such as error messages, stack traces, and multi-component system logs to identify root causes.\n
  • Ingestion points: System logs, stack traces, and test failure messages are explicitly cited as primary data sources in SKILL.md and root-cause-tracing.md.\n
  • Boundary markers: There are no instructions for using delimiters or boundary markers to isolate untrusted error data from the agent's core instructions.\n
  • Capability inventory: The skill possesses significant capabilities including filesystem traversal and shell command execution via utility scripts.\n
  • Sanitization: The process does not include steps for sanitizing or escaping content within processed logs before the agent analyzes them.\n- [PROMPT_INJECTION]: The skill employs strong imperative constraints (e.g., 'The Iron Law', 'NO FIXES WITHOUT ROOT CAUSE', 'ALWAYS', 'NEVER') to strictly govern and override default agent behavior, ensuring adherence to the prescribed debugging methodology.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 1, 2026, 12:55 PM