The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which could contain instructions meant to influence the agent's behavior.
Ingestion points: The skill reads diff content from git diff, commit messages from git log, and specification details via gh issue view or local files in docs/, specs/, and .scratch/.
Boundary markers: The agent spawns parallel sub-agents to isolate the 'Standards' and 'Spec' contexts, providing architectural separation, but it lacks explicit boundary markers or 'ignore' instructions within the sub-agent prompts to prevent the interpretation of embedded data as commands.
Capability inventory: The skill possesses the ability to execute shell commands (git, gh) and read files throughout the repository.
Sanitization: There is no evidence of sanitization, escaping, or validation of the content retrieved from git history or specifications before it is passed to the sub-agents.
[DYNAMIC_EXECUTION]: The skill dynamically generates and executes prompts for sub-agents by interpolating diff fragments and specification text into template instructions.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its core functionality, specifically using git diff, git log, git rev-parse, and gh issue view. It includes a validation step using git rev-parse to ensure the user-supplied reference point is valid before proceeding with more complex operations.

code-review