The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by ingesting untrusted data from the project's code and documentation to generate outputs. * Ingestion points: Data is read from files and Git history based on the scan_paths setting in modules/configuration.md (defaulting to src/, commands/, skills/, and agents/). * Boundary markers: No explicit delimiters are documented to separate scanned content from the agent's internal instructions or to prevent the agent from following instructions found in the codebase. * Capability inventory: The skill uses the gh CLI for issue creation and can write to local files (backlog). * Sanitization: The issue_template in modules/configuration.md interpolates variables like {{ description }} and {{ tradeoffs_summary }} directly into prompts without evidence of escaping, validation, or sanitization.
[COMMAND_EXECUTION]: The skill documentation instructs users to run make test-feature-review, which executes shell commands defined in a local Makefile for logic verification.

feature-review