The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs aggressive, non-negotiable instructional language intended to override the agent's standard reasoning and autonomy.
Evidence: The file contains explicit override markers such as '', 'YOU DO NOT HAVE A CHOICE', 'YOU MUST USE IT', 'This is not negotiable', and 'You cannot rationalize your way out of this'.
Context: These patterns are used to force the agent to invoke skills even when it might otherwise deem them unnecessary, effectively attempting to bypass the model's internal task-relevance filtering.
[INDIRECT_PROMPT_INJECTION]: The skill defines an orchestration workflow that ingests and processes untrusted user input to determine the agent's high-level operation mode.
Ingestion points: User-provided instructions (e.g., 'Add X' or 'Fix Y') are processed in SKILL.md to route the agent between 'Inline' and 'Spec-backed' plans.
Boundary markers: While the skill defines distinct planning modes, it lacks explicit delimiters or instructions to ignore potential commands embedded within the user-supplied task descriptions.
Capability inventory: The orchestrator manages high-privilege activities including file system modifications, architectural changes, and delegation to subagents (via 'spec-implement' and 'code-subagents').
Sanitization: There is no evidence of input validation or sanitization of user instructions before they are used to influence the workflow state.

atelier-orchestrator