The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates external data into the agent's context without sanitization or protective delimiters.
Ingestion points: The skill reads content from local files including production/review-mode.txt, design/live-ops/economy-rules.md, and design/live-ops/ethics-policy.md, and processes user-supplied seasonal descriptions.
Boundary markers: The instructions do not specify any delimiters or warnings to the subagents to ignore potential instructions embedded within the ingested project files.
Capability inventory: The skill possesses significant capabilities, including the ability to spawn subagents via the Task tool and execute commands using Bash, Write, and Edit tools.
Sanitization: There is no evidence of validation, filtering, or escaping of the content read from the filesystem or provided by the user before it is passed to subagents.

team-live-ops