The Agent Skills Directory

[SAFE]: The skill implements a four-phase methodology (Root Cause, Pattern Analysis, Hypothesis, and Implementation) for resolving technical bugs.
[COMMAND_EXECUTION]: The skill includes find-polluter.sh, a local bash script designed to identify tests that cause environmental pollution by executing npm test. This utility operates entirely within the local development environment and does not fetch external payloads.
[DATA_EXFILTRATION]: Diagnostic code examples in SKILL.md use non-leaking shell patterns (e.g., using parameter expansion to check for variable existence) to verify the presence of environment variables without printing their contents to logs or standard output.
[PROMPT_INJECTION]: Included markdown files starting with 'test-pressure-' are evaluation sets designed to assess the agent's decision-making consistency within the skill's framework under simulated constraints. These are self-contained scenarios and are not malicious instructions meant to subvert the agent's core safety logic.

systematic-debugging