The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill's core function is to ingest untrusted external data in the form of Terraform (.tf) files.
Ingestion points: Reads all .tf files within a user-specified directory.
Boundary markers: None visible in documentation; the skill extracts resource names and dependencies which are human-authored strings.
Capability inventory: Accesses the local file system and produces structured diagram code (Mermaid/PlantUML) often interpreted by other agent components.
Sanitization: No evidence of sanitization for malicious strings embedded in resource metadata or comments.
Data Exposure (MEDIUM): Terraform files are sensitive as they define organizational infrastructure.
Evidence: The skill scans entire directories for .tf files. These files often contain internal IP ranges, resource naming schemes, and occasionally hardcoded secrets or environment variables.
Self-referential Safety Claim: The 'Knowledge Protocol' mentioned in SKILL.md claims to prevent leaks of confidential data; however, this is an unverified claim that cannot be validated without the implementation code.

terraform-arch-mapper