The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands for version control operations, including git diff, git diff --cached, git diff HEAD, and git branch. It also utilizes the GitHub CLI (gh pr view) to resolve pull request references.
[COMMAND_EXECUTION]: The workflow architecture executes arbitrary commands defined in the workflow.activation_steps_prepend and workflow.activation_steps_append configuration arrays loaded from customize.toml and project-specific override files.
[REMOTE_CODE_EXECUTION]: The skill uses uv run to execute a script located at {project-root}/_bmad/scripts/resolve_customization.py. This script is external to the skill's distribution and resides within the user's project directory under the vendor's _bmad namespace.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data during its core workflow, creating a vulnerability surface for malicious instructions embedded in project files or code changes.
Ingestion points: Data is ingested from git diffs (diff_output), specification/story files ({spec_file}), and sprint tracking documents ({sprint_status}).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when passing the diff_output to sub-agents like the Blind Hunter or Edge Case Hunter.
Capability inventory: The skill possesses capabilities to execute shell commands (git, gh, uv), perform network reads via GitHub CLI, and write findings back to project files and backlog templates.
Sanitization: No explicit sanitization, escaping, or validation of the content of code diffs or specification files is performed before they are processed by the LLM or sub-agents.

gds-code-review