The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Bun.spawn to execute local scripts and system git commands to perform its versioning tasks.
Evidence: scripts/git-commit.mjs executes git add and git commit to save version changes.
Evidence: scripts/git-push.mjs executes git push to synchronize changes with a remote repository.
Evidence: scripts/version-apply.mjs spawns version-find.mjs as a subprocess to identify target files.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes the contents of project manifest files, which could contain malicious instructions embedded in metadata fields.
Ingestion points: The scripts version-find.mjs, version-current.mjs, and version-apply.mjs read content from package.json, Cargo.toml, pyproject.toml, setup.py, and other code files.
Boundary markers: There are no explicit delimiters or warnings defined to prevent the agent from following instructions that might be contained within the read files.
Capability inventory: The agent has the authority to write to the local filesystem and execute git push and git commit via the provided scripts.
Sanitization: While the scripts use strict semantic versioning regex patterns to validate the versions they write, the data read from files and passed to the agent context is not sanitized for natural language instructions.
[DATA_EXFILTRATION]: The skill provides a direct mechanism to send local data to a remote repository via the Git push helper.
Evidence: scripts/git-push.mjs performs network operations to origin. This is a core functionality of the skill but represents a controlled exfiltration path. The script includes a safety check to prevent force-pushing to main or master branches.

version-update