The Agent Skills Directory

[SAFE]: The skill defines a rigorous structural workflow for planning software development tasks. A key component of this workflow is the mandatory "Plan Review Army" which requires specific evaluations from specialized personas for engineering, CEO, design, and security perspectives before any plan is approved.
[COMMAND_EXECUTION]: The workflow and its associated templates include the generation of shell commands (e.g., curl, npm test) for verifying task completion. These are used within the context of a developer-led testing process and are presented as illustrative examples for user-approved implementation workflows.
[PROMPT_INJECTION]: The skill processes feature specifications and design documents, creating a surface for indirect prompt injection attacks where malicious instructions could be embedded in the project documentation.
Ingestion points: Reads 01-spec.md and design files in the docs/features/ directory to generate implementation plans.
Boundary markers: No specific text delimiters or isolation techniques are mentioned to differentiate user specifications from planner instructions.
Capability inventory: The planner has the capability to write task plans to the filesystem and define commands for subsequent execution by other agents.
Sanitization: While there is no automated filtering of the input documentation, the skill mandates a multi-layered review by specialized agents and final manual approval by the user, providing a significant barrier against malicious instruction following.

build-workflow-plan