The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes execSync and shell scripts to manage database operations and trajectory tracking.
data-migration-example.js and trajectory-optimization-example.js execute local Python and Bash scripts (migration-tool.py, trajectory-tracker.sh).
Examples in SKILL.md suggest using npx agentdb@latest to initialize databases and start MCP servers, which involves downloading and executing remote packages.
[INDIRECT_PROMPT_INJECTION]: The skill implements adaptive learning by storing, retrieving, and synthesizing "experiences" and "trajectories." This creates a vulnerability surface where malicious instructions embedded in stored data could influence the agent's behavior during memory distillation or context synthesis.
Ingestion points: trajectory-tracker.sh (via command-line arguments for task results) and the insertPattern API in SKILL.md.
Boundary markers: The provided scripts do not implement explicit delimiters or "ignore embedded instructions" headers when synthesizing memory context.
Capability inventory: The agent has access to Bash, execSync, and file system operations across the skill's environment.
Sanitization: While the skill documentation recommends sanitizing metadata, the implementation in trajectory-tracker.sh relies on jq for JSON escaping, which ensures structural integrity but does not perform semantic sanitization of potential prompt injection content.
[EXTERNAL_DOWNLOADS]: The skill instructions and scripts download the agentdb package from the NPM registry via npx. This is documented as a standard dependency but represents an external code fetch.
[DYNAMIC_EXECUTION]: The skill dynamically generates and executes logic across multiple environments, including Node.js, Python, and Bash, often passing data between them via command-line arguments and JSON files.

reasoningbank-with-agentdb