The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied project details and organizational information to generate various reports and plans, creating a surface for indirect prompt injection.
Ingestion points: The agent takes direct user input for project goals, stakeholder details, and strategic requirements as described in SKILL.md and references/project-tracker.md.
Boundary markers: The instructions do not define specific delimiters (like triple-backticks or unique separators) or provide "ignore embedded instructions" warnings for the agent to use when processing this external data.
Capability inventory: The skill workflow in references/project-tracker.md includes the capability to write multiple files (e.g., 01_project_plan.md, 02_milestone_tracking.md) to a local _workspace/ directory.
Sanitization: There are no defined sanitization, validation, or filtering mechanisms mentioned to handle potentially malicious or malformed input from the user before it is incorporated into the generated documents.

collab-roadmap