The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the execution environment.
Ingestion points: The skill explicitly identifies stderr_tail and Failed test output as primary inputs for its classification logic in SKILL.md.
Boundary markers: There are no boundary markers or instructions to the agent to disregard potential instructions embedded within the stderr or test output data.
Capability inventory: The skill utilizes Read, Grep, Glob, and Bash to interact with the file system and process data.
Sanitization: The instructions do not define any sanitization, escaping, or filtering procedures for the stderr content before it is matched against regex patterns or evaluated by the rules.
[COMMAND_EXECUTION]: The skill requires the Bash tool to execute its classification logic. While the described use case (checking exit codes and matching strings in logs) is standard, the use of a shell environment to process untrusted log data creates a potential execution risk if shell metacharacters are not properly handled during the evaluation of the rules.

failure-classifier