The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external codebase content, which presents a surface for instructions embedded in data (such as code comments) to potentially influence the agent's behavior during analysis.
Ingestion points: Project source code across multiple languages (C++, Go, Java, JavaScript, Python) in files like resources/codeql/python.ql and resources/semgrep/cpp.yaml that target user-controlled data sources.
Boundary markers: None explicitly defined to separate code content from instructions during the triage and reading phases.
Capability inventory: File system read access, execution of search tools (ripgrep, semgrep, CodeQL), and report generation capabilities.
Sanitization: The agent is instructed to read raw code snippets directly for analysis without intermediate filtering.
[COMMAND_EXECUTION]: The skill workflow requires the agent to execute standard command-line utilities to perform searches and analysis.
Evidence: The instructions in references/searching.md and SKILL.md detail the use of ripgrep (rg), semgrep, and CodeQL to identify vulnerability patterns in the codebase.

variant-analysis