The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands for repository management and software testing.\n
Evidence: references/spec-merge.md includes commands such as git merge, git branch -d, and git worktree remove.\n
Evidence: references/spec-go.md and references/spec-task.md instruct the agent to execute test and lint commands specified within the design.md file of the specification directory.\n- [PROMPT_INJECTION]: The skill's implementation loop is vulnerable to indirect prompt injection through the processing of user-controlled or project-resident markdown files.\n
Ingestion points: requirements.md, design.md, and tasks.md in the specification directory, as well as files in the steering/ directory.\n
Boundary markers: There are no explicit delimiters or instructions to the agent to disregard commands embedded within the specification content.\n
Capability inventory: The agent has the capability to read/write files, perform git operations, and execute shell commands for testing and linting.\n
Sanitization: The skill does not implement validation or sanitization of the content provided in the specification files before acting on it.

spec-driven-dev