The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes find-polluter.sh, a bash utility that identifies which tests in a project are creating unwanted persistent state (pollution) by executing npm test on files discovered via a user-provided pattern. This is a standard development tool and does not present an inherent security risk in this context.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and analyzing error logs, stack traces, and system output. This represents a surface for indirect prompt injection if an attacker can control log content. However, the skill's mandatory 4-phase process (requiring reproduction, hypothesis forming, and minimal testing) acts as a procedural safeguard against executing commands based solely on untrusted log strings.
[SAFE]: Diagnostic examples provided in the SKILL.md file demonstrate secure logging practices, such as checking for the presence of secrets using ${IDENTITY:+SET} syntax to verify environment configuration without revealing the actual values of those secrets.
[SAFE]: The skill uses 'Pressure Tests' (markdown scenarios) to train the agent to resist shortcuts or social pressure to bypass its core debugging principles. These are behavioral alignment tools rather than malicious prompt injections.

systematic-debugging