The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/apply.py and scripts/scaffold.py to initialize a Git repository (git init, git add, git commit). These operations are used solely to set up version control for the newly created workspace and are part of the intended functionality.
[COMMAND_EXECUTION]: The SKILL.md instructions suggest running python3 -m http.server 8899 to serve a local HTML plan report. This allows the user to review the proposed file changes in a web browser before execution.
[EXTERNAL_DOWNLOADS]: The documentation references the official GitHub repository for the tool (github.com/SpaceZephyr/build-your-harness) as a source for installation and reference. This follows the vendor resource pattern and is considered a safe reference to the skill's own origin.
[INDIRECT_PROMPT_INJECTION]: The skill includes functionality to ingest local file metadata (filenames and paths) to categorize items into the workspace structure.
Ingestion points: scripts/migrate.py and scripts/plan.py scan directory structures from user-provided source paths.
Boundary markers: The skill enforces a 'Plan-Review-Apply' workflow, requiring the user to manually copy a decision JSON from an HTML report back into the chat, serving as a robust human-in-the-loop checkpoint.
Capability inventory: The skill has capabilities for file system writing, Git command execution, and starting a local HTTP server.
Sanitization: The scripts/plan.py script uses html.escape to sanitize file paths and names before rendering them in the HTML report, mitigating potential script injection in the review UI.

star-your-harness