The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes and executes a local Python script (_system/scripts/memory_manager.py) to manage and filter its internal memory system (learnings.jsonl). This script performs data processing tasks such as worker-based filtering, deduplication, and confidence-based decay of session insights.
[PROMPT_INJECTION]: The skill architecture creates a surface for indirect prompt injection by ingesting and storing untrusted data from external sources like competitor websites and market research. Ingestion points include the afa-compete and afa-explore modules, where retrieved content is stored in Markdown and JSONL files within the ./brand-brain/ directory. The system lacks explicit instruction-ignore delimiters or robust sanitization for this external content before it is retrieved for prompt interpolation. Given the agent's significant capabilities in generating marketing assets and managing files, malicious instructions embedded in external data could potentially influence agent behavior.

afa