The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file references/08-docker.md contains a hardcoded connection string postgresql://postgres:password@db:5432/mydb within a Docker Compose configuration example. While this is a common placeholder pattern for local development documentation, it matches credential exposure signatures.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process monorepo configuration files and source code which could be attacker-controlled in shared environments.
Ingestion points: The skill reads turbo.json, package.json, and source files across the workspace to determine task graphs and hashing.
Boundary markers: No explicit instructions are provided to the agent to ignore or delimit potentially malicious instructions embedded within the monorepo files it processes.
Capability inventory: The skill facilitates extensive shell command execution via the turbo CLI, including turbo run, turbo prune, and turbo watch, which execute scripts defined in package.json files.
Sanitization: There is no evidence of input sanitization or validation for the content of the scripts or configuration files before execution by the agent.

turborepo