The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process session transcripts and logs, which are external data sources that may contain untrusted content.\n
Ingestion points: The agent is instructed to review logs and transcripts identified by the scripts/dream_memory.py script as specified in references/prompt-template.md.\n
Boundary markers: The prompt template uses placeholders like <memory_report> to structure input, but lacks explicit guardrails or delimiters to prevent the agent from executing instructions potentially hidden within the processed logs.\n
Capability inventory: The skill allows the agent to read and write files within the memory and project directories to update topic files and MEMORY.md.\n
Sanitization: No content filtering or sanitization of the logs is performed before the agent processes them.\n- [COMMAND_EXECUTION]: The skill executes a local Python script scripts/dream_memory.py to generate reports on the state of the memory directory.\n
Evidence: SKILL.md contains the command python3 {baseDir}/scripts/dream_memory.py --memory-root /path/to/memory --transcripts-dir /path/to/transcripts.\n
Analysis: The script is part of the skill package and its behavior is limited to file system metadata analysis (size, modification time) and indexing within specified paths.

dream-memory