The Agent Skills Directory

[COMMAND_EXECUTION]: The skill incorporates local shell scripts (scripts/generate-changelog.sh, scripts/next-version.sh, and scripts/validate-commits.sh) to automate Git-related tasks. These scripts primarily perform read-only operations on the repository history using git log and git describe to parse commit metadata. The logic is confined to repository management and does not involve risky system-level commands or unauthorized access.
[DATA_EXFILTRATION]: No network operations or external data transfer mechanisms were found. On the contrary, the skill includes a security rule specifically designed to prevent data exposure by auditing staged changes for sensitive patterns (e.g., .env, credentials.*, *.pem) and warning the user before commits are made.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (commit messages from the git history) to generate changelogs.
Ingestion points: The scripts ingest commit messages via git log command output.
Boundary markers: No explicit delimiters are used within the shell scripts for the ingested text.
Capability inventory: The skill facilitates git commit, git tag, and git push operations.
Sanitization: Commit messages are processed through sed for formatting into sections, but no explicit sanitization for injection is present. While this represents a theoretical attack surface for indirect injection via malicious commit messages, the risk is assessed as safe given the narrow technical scope of the transformation logic and the intended developer-focused usage environment.

git-workflow-automation