The Agent Skills Directory

**[Prompt Injection
Indirect] (HIGH):** The skill's core purpose is to read, parse, and execute instructions from an external 'implementation plan' file. This creates a significant attack surface where a malicious plan could override the agent's instructions.
Ingestion points: Implementation plan files (referenced in description and Phase 1).
Boundary markers: Absent. The instructions command the agent to 'Read and parse the implementation plan completely' without delimiters or warnings to ignore embedded instructions.
Capability inventory: Bash, Write, Edit, Task, TodoWrite (powerful tools for system modification).
Sanitization: Absent. There is no logic to filter or validate the content of the plan before performing impact analysis or 'Executing planned updates'.
[Command Execution] (MEDIUM): The skill explicitly grants the Bash tool. While intended for discovery (Grep, Glob), the lack of constraints means any instruction found within a malicious 'Implementation Plan' could lead to arbitrary command execution on the host system.

plan-analysis