The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external SARIF files, which serves as an ingestion point for potentially untrusted data derived from various security scanners.
Ingestion points: Results files (e.g., results.sarif) are parsed using standard JSON libraries or streamed via ijson in SKILL.md and sarif_helpers.py.
Boundary markers: The skill leverages the structured nature of the SARIF JSON schema to maintain data boundaries, although result messages are extracted for display/analysis.
Capability inventory: The skill uses tools like Bash, Read, Glob, and Grep. Python scripts perform file operations and data aggregation.
Sanitization: Content is processed into structured Finding objects; while explicit filtering of malicious text within message fields is not present, the reliance on structured data parsing reduces risk.
[EXTERNAL_DOWNLOADS]: The instructions recommend the installation of specialized tools and libraries for SARIF manipulation.
Python Packages: Suggests pysarif, sarif-tools, ijson, and jsonschema. The sarif-tools package is maintained by Microsoft's official repository.
Node.js Packages: Mentions ajv-cli for schema validation.
Other Utilities: References standard tools like jq and the official SARIF validator at sarifweb.azurewebsites.net.
[COMMAND_EXECUTION]: The skill provides jq and bash command examples for filtering and aggregating scan results. These commands are standard for data processing tasks and do not involve administrative privileges or dangerous shell operations.

sarif-parsing