The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from the local repository which creates a potential injection surface.
Ingestion points: The skill instructions direct the agent to read project authority docs, recent commit history, and content from CONTEXT.md files (found in SKILL.md).
Boundary markers: The skill does not define explicit delimiters or include instructions to disregard potential commands embedded within the ingested project files.
Capability inventory: Across the provided files, the skill has the capability to write design documents and specs to the docs/aegis/ directory and execute shell-based Python commands (found in expanded-design-guidance.md).
Sanitization: There is no evidence of sanitization or filtering logic applied to the ingested project context before it is used to inform the agent's design recommendations.
[COMMAND_EXECUTION]: The skill utilizes a local Python helper utility (aegis-workspace-helper) to initialize the workspace and manage documentation indices. The commands include parameters such as --root and --path which are derived from the local file system and project structure (found in expanded-design-guidance.md).

brainstorming