The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and analyze external security scan findings, which could potentially contain malicious payloads or instructions aimed at influencing the agent's reasoning.
Ingestion points: The skill retrieves threat models and security candidates from artifact locations, such as ../../references/scan-artifacts.md and candidate lists.
Boundary markers: The skill does not explicitly define structural delimiters for the data it processes, instead relying on instructional logic to maintain analysis focus.
Capability inventory: The skill utilizes tools like list_codex_security_candidates and record_candidate_attack_paths to interact with scan data and record analysis outcomes.
Sanitization: The skill emphasizes rigorous validation through counterevidence checklists and severity policy calibration, which acts as a logical control when processing potentially untrusted data.

attack-path-analysis