The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes SARIF files which are untrusted external data sources, creating a vulnerability surface for indirect prompt injection. Malicious instructions embedded in security findings could attempt to manipulate the agent. \n
Ingestion points: The skill uses jq commands in SKILL.md and the load_sarif function in resources/sarif_helpers.py to read external files. \n
Boundary markers: There are no explicit markers or instructions to the agent to disregard embedded directives within the SARIF content. \n
Capability inventory: The skill has access to Bash, Read, Glob, and Grep tools, which could be exploited by a successful injection. \n
Sanitization: No sanitization is performed on the text fields of the SARIF results (such as messages or rule descriptions) before they are processed by the agent. \n- [EXTERNAL_DOWNLOADS]: The skill documentation and helper scripts reference multiple external packages and repositories for extended functionality. \n
Documentation suggests the use of Python packages pysarif, ijson, and jsonschema. \n
References the garif Go package hosted on GitHub at github.com/chavacava/garif. \n
Mentions the ajv-cli Node.js package for validation purposes. \n
References the sarifweb.azurewebsites.net service for online validation. \n
References various repositories including github.com/Kjeld-P/pysarif and Microsoft's sarif-tools.

sarif-parsing