The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon instructions defined in external files like plan.md and gate.md. A malicious actor who can modify these files could embed instructions that the agent might execute as part of the 'Task Execution' phase.
Ingestion points: The agent is instructed to read .beyond-code/<slug>/gate.md and .beyond-code/<slug>/plan.md in Stage 0 and Stage 1.
Boundary markers: There are no specified delimiters or instructions to treat the content of these files as untrusted data, increasing the risk that the AI will follow instructions embedded within the plan.
Capability inventory: The skill allows for file modifications, task execution, and shell command execution (specifically git log, git add, and git commit).
Sanitization: The instructions do not define any sanitization or validation logic for the content of the tasks being executed beyond topological ordering.

beyond-code-build