The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform environment checks, manipulate local YAML files using yq, and run a TypeScript-based consistency checker using bun.
[REMOTE_CODE_EXECUTION]: Internal shell scripts in references/stage-2-naming.md and references/stage-3-entities.md interpolate variables (e.g., ${NAMING_MODE_VALUE}, ${SLUG}) derived from user input or processed content directly into command strings without sanitization.
[EXTERNAL_DOWNLOADS]: The skill implemented a mechanism to fetch blueprint files from remote URLs using curl -s.
[PROMPT_INJECTION]: An indirect prompt injection surface is present. 1. Ingestion points: SKILL.md (loading artifacts) and curl (fetching URLs). 2. Boundary markers: No markers or 'ignore' instructions for ingested file content. 3. Capability inventory: High-privilege shell access via Bash and bun. 4. Sanitization: Absent for ingested data before processing. This creates a scenario where a malicious blueprint file could influence the agent's high-privilege operations.

semantius-architect