The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (HIGH): The skill is designed to execute arbitrary code within the target project's environment using npm test. If an agent uses this skill on a malicious or compromised repository, it will execute untrusted code without any sandbox or safety constraints.
Indirect Prompt Injection (HIGH): This skill provides a significant attack surface for indirect prompt injection as it requires the agent to deeply analyze and act upon untrusted external data (source code, logs, and error messages).
Ingestion points: Filenames processed by scripts/find-polluter.sh and the source code/error logs analyzed in references/systematic-debugging.md.
Boundary markers: Absent. There are no instructions for the agent to use delimiters or to ignore instructions embedded within the code it is debugging.
Capability inventory: High-privilege capabilities including the ability to execute shell scripts and run package-manager-level test commands (npm test).
Sanitization: Absent. There is no evidence of sanitization, validation, or escaping of the external content before it is processed or executed.
Dynamic Execution (MEDIUM): The scripts/find-polluter.sh script employs unsafe shell practices. Specifically, the loop for TEST_FILE in $TEST_FILES relies on unquoted variable expansion of output from the find command. This makes the script vulnerable to word-splitting and potential command injection if a repository contains filenames with shell metacharacters or spaces.

debugging