The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a persistent learning feedback loop where agent observations are recorded in references/learnings.md and re-ingested as context in future sessions.
Ingestion points: Untrusted user queries and the references/learnings.md file are loaded into the agent context.
Boundary markers: Absent; the instructions do not provide delimiters or safety warnings to distinguish core logic from accumulated 'learnings'.
Capability inventory: The skill is restricted to reading and writing local markdown files; it lacks network access, shell execution, or privilege escalation capabilities.
Sanitization: Absent; there is no validation or filtering of content before it is stored in the persistent learning file.

sales-scribbl