The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes cat to read project files and runs a Node.js script (memplan-cli.js) from a system-defined path to manage and update the project plan state.
[PROMPT_INJECTION]: Indirect Prompt Injection: The skill processes user-controlled data from .memplan/plan.mem and .memplan/progress. This creates a potential surface for indirect injection where data in these files could influence agent logic.
Ingestion points: Reads .memplan/plan.mem and .memplan/progress files.
Boundary markers: None identified in the parsing logic.
Capability inventory: Shell command execution (cat) and file write/update capabilities via the memplan-cli.js tool.
Sanitization: No explicit content validation or sanitization is observed in the processing of the plan entries.

refine