The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting untrusted incident descriptions, event logs, and metadata to generate structured reports and communication templates. \n- Ingestion points: All core scripts, including scripts/incident_classifier.py and scripts/pir_generator.py, read user-provided incident data from files or standard input. \n- Boundary markers: The templates used for generating communications (found in references/communication_templates.md and SKILL.md) use standard Python string formatting (e.g., .format()) to insert user data. There is no evidence of specific boundary delimiters or instructions to ignore embedded commands within the input data. \n- Capability inventory: The scripts are restricted to standard data processing and writing to local files via CLI arguments. No network operations, arbitrary command execution, or sensitive file system access were identified in the source code. \n- Sanitization: The analysis did not identify any logic to sanitize or filter user-supplied text for potential prompt injection patterns before it is processed by the AI agent or included in generated documents.

incident-commander