The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses strict instructional language such as "You MUST" and "Do NOT invoke" to enforce a design-before-implementation workflow. These are logical constraints intended to guide the agent's behavior rather than malicious bypass attempts.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests data from the current project environment, creating a potential surface for indirect prompt injection.
Ingestion points: The agent is directed to explore project files, documentation, and recent commit history to understand context (SKILL.md).
Boundary markers: There are no specified delimiters or instructions to treat existing file content as untrusted or to ignore instructions embedded within those files.
Capability inventory: The skill has the capability to write new documentation files to the docs/plans/ directory, perform git commits, and transition execution to the writing-plans skill (SKILL.md).
Sanitization: No validation or sanitization of the content read from the project files is mentioned before it is used to influence the design process.

brainstorming