The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and analyze data from external security scans and threat models, which represents an attack surface for indirect prompt injection. * Ingestion points: Data is ingested through the list_codex_security_candidates tool and from external scan artifact files as defined in scan-artifacts.md. * Boundary markers: The instructions do not define specific boundary markers or delimiters to separate the agent's internal reasoning instructions from the untrusted scan data it analyzes. * Capability inventory: The skill possesses the capability to write to the file system (attack-path receipts) and update scan candidate records using the record_candidate_attack_paths tool. * Sanitization: No explicit sanitization, filtering, or validation steps are defined for the ingested scan findings before they are processed by the agent. * Context: As a security analysis tool, it is expected to handle potentially deceptive or malicious content. However, the lack of explicit boundaries between the instructions and the analyzed data could theoretically allow a sophisticated payload within a scanned finding to influence the agent's severity calibration or reportability decisions.

attack-path-analysis