The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to systematically read and execute instructions from todo-list.md and plan.md files. This creates a surface for indirect prompt injection if these files are provided by untrusted sources or contains malicious instructions embedded in the project specs.
Ingestion points: specs/active/[task-id]/plan.md, spec.md, tasks.md, and todo-list.md.
Boundary markers: Absent. The agent is instructed to 'Read entire list before starting' and 'Execute in order' without delimiters to separate instructions from data.
Capability inventory: The agent is authorized to generate code, modify the filesystem, and execute the local scripts/progress.sh utility.
Sanitization: None. There is no validation or filtering of the content within the task files before the agent attempts to implement the described features.
[Command Execution] (SAFE): The skill includes a local utility scripts/progress.sh. Analysis of the script shows it is safe; it uses find and grep with proper quoting to parse local file statuses and does not execute untrusted input or perform network operations.

sdd-implementation