The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Node.js wrapper (scripts/main.cjs) to execute Git commands via the child_process.spawn method.
The implementation uses shell: false and windowsHide: true, which are security best practices that mitigate shell injection and unauthorized visibility on the host system.
It passes command-line arguments directly to the Git binary, allowing the agent to perform comprehensive repository management tasks.
[PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection (Category 8) due to its core function of reading and processing external repository data.
Ingestion points: Untrusted data enters the agent's context through Git operations like git log, git show, and git diff, which are executed by scripts/main.cjs.
Boundary markers: The skill does not specify any boundary markers or instructions for the agent to ignore or isolate instructions that might be embedded in Git commit messages or file content.
Capability inventory: The skill has the capability to execute subprocesses (via scripts/main.cjs) and is instructed by the 'Memory Protocol' in SKILL.md to write findings and decisions to the local file system (e.g., .claude/context/memory/learnings.md).
Sanitization: There is no evidence of sanitization or validation of the data retrieved from the Git repository before it is processed or written to the agent's internal memory files.

git-expert