The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill architecture involves ingesting untrusted external data which is then used to prompt sub-agents with high-privilege capabilities.
Ingestion points: The skill accepts external file paths, descriptions, and reference files as inputs in Step 0.
Boundary markers: Step 5's prompt template does not utilize boundary markers or 'ignore' instructions to isolate ingested content from agent instructions.
Capability inventory: The system grants sub-agents full access to tools including 'Read', 'Write', and 'Shell' as described in the 'Agents' section.
Sanitization: There is no evidence of sanitization or safety-filtering for the plan inputs before they are interpolated into the sub-agent prompts.
[COMMAND_EXECUTION]: The skill utilizes the shell to perform project-level verification tasks.
Evidence: Step 6.3 specifies the execution of npm run scripts for type checking, linting, and testing, which grants the agent the ability to execute code defined within the project environment.

parallel-builder