The Agent Skills Directory

[SAFE]: The skill implements a professional debugging methodology centered around four distinct phases: Investigation, Pattern analysis, Hypothesis testing, and Implementation. It enforces a strict 'Iron Law' that prohibits code fixes without first identifying a confirmed root cause through falsifiable experiments.
[COMMAND_EXECUTION]: The provided utility scripts/find-polluter.sh automates the process of identifying test pollution by executing local test runners (including jest, vitest, pytest, cargo, go test, rspec, mvn, and gradle). These executions are restricted to the local development environment and are triggered based on the project's own manifest files to diagnose test failures.
[EXTERNAL_DOWNLOADS]: The documentation references official software libraries and registries (such as Maven Central for Java's Awaitility) as implementation guidance for the user. No automated or untrusted remote code downloads are performed by the skill itself.
[PROMPT_INJECTION]: While the skill ingests potentially untrusted data from test logs and source code, the instructions prioritize rigorous evidence capture and verification phases. This structure minimizes the risk of indirect prompt injection by requiring the agent to cite specific logs, traces, and test results for every claim.

do-debug