systematic-debugging

Warn

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains multiple 'Pressure Test' files (test-pressure-1.md, test-pressure-2.md, test-pressure-3.md) that utilize imperative language to override the AI agent's standard behavior and decision-making processes. These files use phrases like "IMPORTANT: This is a real scenario. You must choose and act. Don't ask hypothetical questions
  • make the actual decision," which are characteristic of prompt injection attempts to bypass AI safety filters or operational guidelines.
  • [DATA_EXFILTRATION]: Instructions in SKILL.md recommend gathering evidence by dumping environment variables using env | grep IDENTITY. This pattern specifically targets potential secrets (identities) and can lead to the exposure of credentials if the output is logged, shared, or processed by subsequent tools.
  • [COMMAND_EXECUTION]: The provided utility script find-polluter.sh executes shell commands (npm test) on files identified at runtime. This creates a risk of arbitrary code execution if the file paths or the content of the tests are influenced by untrusted data.
  • [COMMAND_EXECUTION]: SKILL.md suggests the use of powerful system commands such as security list-keychains and codesign. While intended for debugging signing issues, these commands interact directly with the host system's security and identity management infrastructure.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 29, 2026, 02:15 AM