The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to perform shell commands to manage the local development environment and verify code quality.
It executes Git commands such as git checkout -b and git worktree add for branching and worktree creation.
It identifies and runs project-specific testing, linting, and formatting tools (e.g., npm test, pytest) identified from local configuration files during the orientation phase.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection through its integration with external issue tracking data.
Ingestion points: Section 1 ('Orient') of SKILL.md instructs the agent to fetch external data using gh issue view <number> --comments to determine the behavior sequence.
Boundary markers: The skill does not employ delimiters or specific instructions to the agent to treat the retrieved GitHub content as untrusted data rather than executable instructions.
Capability inventory: The agent has permission to modify the file system, create Git commits, and execute arbitrary shell commands for testing purposes across all scripts.
Sanitization: There is no specified mechanism to filter, validate, or sanitize the content retrieved from GitHub before it is incorporated into the agent's context.

snap-forge