The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates an 'indirect prompt injection' surface through its 'skill-context' mechanism. It is explicitly instructed to read and treat instructions from an external file (.ai-factory/skill-context/aif-roadmap/SKILL.md) as mandatory overrides that 'win' over the skill's default rules.
Ingestion points: The agent is directed to read .ai-factory/skill-context/aif-roadmap/SKILL.md, .ai-factory/DESCRIPTION.md, and .ai-factory/ARCHITECTURE.md to determine its operating logic.
Boundary markers: Absent. There are no delimiters or instructions to treat the ingested content as data rather than instructions; conversely, the agent is told 'the skill-context rule wins' and 'Do NOT ignore skill-context rules'.
Capability inventory: The skill possesses capabilities to Write, Edit, and execute Bash(git *) commands, which could be misused if the context files are manipulated.
Sanitization: None. The skill does not validate or sanitize the rules imported from the workspace before adopting them as its primary operational logic.

aif-roadmap