The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses specific directives (e.g., "IMPORTANT: do NOT delegate") to manage multi-agent orchestration. These are functional instructions for the skill's coordination logic rather than attempts to bypass platform safety filters.
[PROMPT_INJECTION]: The skill processes untrusted external data from sources such as _user-data.md and project documentation, presenting an indirect prompt injection surface. This is a characteristic risk of the skill's intended analysis function.
Ingestion points: Reads docs/roadmap/, docs/specs/, docs/architecture/, and docs/sales-plans/_user-data.md.
Boundary markers: None explicitly defined for input data interpolation within the prompt templates.
Capability inventory: Employs TeamCreate, TaskCreate, and Agent tools; spawned agents have file system access via role definitions.
Sanitization: No specific sanitization or validation of external file content is described before interpolation.
[DATA_EXFILTRATION]: The skill contains a "Shared Principles" section that explicitly forbids the inclusion of credentials, API keys, and PII in any agent communication or context files, which is a significant positive security control.

plan-sales