The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by instructing the agent to execute local Python scripts (scripts/parse_error.py and scripts/analyze_error.py) to process user logs and perform repository searches. These are benign functional commands restricted to the skill's specific purpose.
[DATA_EXPOSURE_AND_EXFILTRATION]: The analyze_error.py script features a repository search capability that traverses a directory provided by the user via the --repo-path argument. It reads files with common code and configuration extensions (.py, .yaml, .json, etc.) to find error-related strings. This facilitates local file system read access for the purpose of debugging source code.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted data from user-provided error logs, stack traces, and environment descriptions via text input or log files.
Boundary markers: The instructions recommend using structured markdown headers (e.g., ## Error Information, ## Environment) to separate and organize input data for the parser.
Capability inventory: The skill possesses file system read access through the search functionality but lacks capabilities for network exfiltration, arbitrary shell command execution, or dynamic code evaluation.
Sanitization: The scripts utilize regex pattern matching and string search operations on the input data without explicit escaping or sanitization of the content.

error-analyzer