The Agent Skills Directory

[COMMAND_EXECUTION]: Multiple scripts, including scripts/install_memory_3layer.py, hooks/memory_paths.py, and testing files, utilize subprocess.run() to execute system commands for Git repository detection and package validation. This is a core part of the skill's ability to locate project roots across different environments.\n- [REMOTE_CODE_EXECUTION]: The skill uses dynamic execution techniques to manage its runtime hooks and self-checks. The install_memory_3layer.py script generates a Python launcher string that uses runpy.run_path() to load and execute hook scripts from dynamically computed paths. Additionally, scripts/check_memory_3layer.py uses the compile() function to validate the syntax of the package's source code.\n- [DATA_EXFILTRATION]: While the skill reads project data to build its memory layers, it implements active sanitization in hooks/memory_extractor.py and scripts/migrate_legacy_memory.py. These scripts use regular expressions to detect and block the recording of sensitive data, such as private keys, API tokens, and machine-specific paths, effectively mitigating unintended data exposure.\n- [PROMPT_INJECTION]: The skill introduces an attack surface for indirect prompt injection by aggregating untrusted data from project files (knowledge base, notes, and facts) into the agent's context via hooks/memory_loader.py. Mandatory evidence:\n
Ingestion points: .memory-3layer/MEMORY.md, .memory-3layer/memory/*.md, and .memory-3layer/areas/topics/*/items.json.\n
Boundary markers: The loader uses Markdown headers and truncated context warnings to delimit memory sections.\n
Capability inventory: The skill can write to the filesystem and execute Git commands via subprocess calls in scripts/install_memory_3layer.py.\n
Sanitization: Content is filtered for secrets using regex patterns before persistence.

memory-3layer