The Agent Skills Directory

[SAFE]: The skill implements a robust memory architecture that isolates user data and configuration within a specific folder ({project-root}/_bmad/memory/bmad-agent-dream-weaver/). It correctly separates sensitive user preferences (like names and languages) into a separate user configuration file intended for local use, following security best practices for credential and identity management.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and processes untrusted dream narratives written by the user.
Ingestion points: Journal entries stored in the journal/ subdirectory are parsed by various analysis modules (dream-interpret.md, pattern-discovery.md).
Boundary markers: The skill specifically requires loading access-boundaries.md upon activation to enforce strict read/write limits to the designated memory folder, preventing unauthorized file access.
Capability inventory: The skill maintains the ability to write journal files, update registry YAMLs, and execute local Python analysis scripts via the command line.
Sanitization: Metadata is processed via safe YAML parsing (yaml.safe_load in the provided scripts), and dream narratives are treated as instructional context for analysis rather than executable code.
[COMMAND_EXECUTION]: The skill invokes several local Python scripts (symbol_stats.py, recall_metrics.py, seed_tracker.py, merge-config.py) to perform data processing. These scripts use standard Python libraries, handle path resolution safely by requiring absolute paths, and do not perform network operations or access files outside the skill's defined scope.

bmad-agent-dream-weaver