The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to utilize standard software development tools for verification and version control. Specifically, it uses bun run tsc for type-checking and bun test for execution of test suites, along with standard git commands for committing incremental progress. These operations are local and consistent with standard coding workflows.
[PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external markdown files (specs/*.md). While this provides a potential surface for indirect prompt injection, the risk is mitigated by a mandatory human-in-the-loop requirement where the agent must present a wave plan for user approval before execution. Ingestion points: specs/*.md. Boundary markers: User approval checkpoint in Phase 2. Capability inventory: Sub-agent spawning, file writing, git commits, and shell command execution via bun. Sanitization: Not explicitly implemented in instructions.

spec-execution