The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external source code, logs, and configuration files via tools like Read, Grep, and Glob. This creates a theoretical surface for indirect prompt injection if the audited materials contain malicious instructions designed to influence the auditing agent.
Ingestion points: Processes local files and source code using Read, Grep, and Glob during the evidence collection and failure mapping phases in SKILL.md.
Boundary markers: The instructions do not explicitly define boundary markers or delimiters to isolate audited content from the agent's instructions.
Capability inventory: The skill utilizes Bash, Write, and Edit tools, allowing it to execute commands and modify files based on audit findings.
Sanitization: No explicit sanitization or validation of content read from external files is described.
[SAFE]: No malicious patterns, such as direct prompt injection, credential exfiltration, obfuscation, or persistence mechanisms, were detected in the skill instructions, scripts, or metadata. The provided Bash commands are benign local searches using rg (ripgrep) for architectural patterns.

agent-architecture-audit