The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external task descriptions within subagent prompts, creating a surface for indirect prompt injection.
Ingestion points: The skill ingests untrusted task text into placeholders within prompts/implementer.md and prompts/reviewer.md.
Boundary markers: The prompts use Markdown headers to delimit task content, but do not include specific instructions for subagents to ignore potentially malicious commands embedded within the task text.
Capability inventory: The subagents are granted capabilities to modify the filesystem, execute tests, and perform git operations such as git commit.
Sanitization: The skill does not implement validation or sanitization of the task text before it is interpolated into prompts.
[COMMAND_EXECUTION]: The skill orchestrates the use of standard development tools for workflow management.
Evidence: The references/execute-loop.md and prompts/reviewer.md instructions require the agent to execute shell-based commands such as git rev-parse HEAD and git diff to track changes and perform reviews.

flow-shared