The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection. It collects user-provided data (project goals, constraints, and descriptions) during the intake phase and persists this information in artifact files like INTAKE.md. These artifacts are subsequently read and used to generate execution plans and code in later phases.
Ingestion points: User input collected during Phase 1 (Intake) across seven project dimensions (Goal, Stack, Platform, Audience, Integrations, Constraints, and Quality).
Boundary markers: The instructions do not define specific delimiters or 'ignore' instructions when writing user-provided content into artifacts or when reading them back into the context.
Capability inventory: The skill utilizes 'Bash', 'Write', and 'Edit' tools in Phase 3 (Execute) to build the project based on the generated plan.
Sanitization: No explicit instructions are provided to sanitize or validate user-provided input before it is interpolated into project artifacts or used to define workstreams.
[COMMAND_EXECUTION]: The skill utilizes the 'Bash' tool to perform operational tasks such as initializing projects, managing dependencies (e.g., 'npm ci', 'pip install'), and running quality gate checks. These operations are a core part of the orchestrator's function, and the risk is mitigated by a mandatory user confirmation step for the execution plan in Phase 2.

conductor