The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and acting upon untrusted external data sources such as PLAN.md files, issues, or PRDs.
Ingestion points: The skill explicitly identifies PLAN.md, issues, and PRDs as the primary data sources for its decision-making loop.
Boundary markers: The instructions lack specific delimiters or guardrails to prevent the agent from following malicious instructions that might be embedded within the project plans being processed.
Capability inventory: The skill is authorized to explore the codebase and modify plans, which could be leveraged if the agent is influenced by adversarial content in the input documents.
Sanitization: There is no mention of sanitizing or validating the content of the plans before the agent uses them to generate and auto-select 'Recommended' options.

auto-grill