The Agent Skills Directory

[COMMAND_EXECUTION]: The skill autonomously executes shell commands to perform repository management and development tasks.
Evidence: Uses git for staging, committing, merging, and resetting progress within the user's project directory.
Evidence: Executes project-defined test suites and linters as part of the implementation and review cycle during /spec-go and /spec-task commands.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes external project data to drive its automated implementation loop.
Ingestion points: The agent reads various Markdown files in the .kiro/specs/ directory (requirements.md, design.md, tasks.md, spec.md), as well as project README.md and steering documents.
Boundary markers: There are no explicit delimiters or specific instructions provided to the agent to disregard instructions embedded within these files.
Capability inventory: The agent has the authority to write files, perform git operations, and execute project-defined code (tests and linters).
Sanitization: The skill does not perform validation or sanitization of the content of the specification files before using them to generate or modify source code.

spec-driven-dev