The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/worktree_tidy.py performs operations using subprocess.run to call system binaries like git, gh, glab, and cnb. These calls are structured as argument lists without shell=True, which mitigates the risk of shell injection.
[DATA_EXFILTRATION]: The skill initiates network requests via CLI tools to external domains (GitHub, GitLab, and cnb.cool) to retrieve pull request metadata. While these are necessary for the skill's functionality, they constitute network activity to domains outside the standard whitelist.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted metadata from Git repositories, such as commit messages, which could be used for indirect prompt injection attacks.
Ingestion points: Commit subjects are extracted from git log and filenames from git status in scripts/worktree_tidy.py.
Boundary markers: No explicit delimiters are used in the script's output to separate untrusted content from instructional output.
Capability inventory: The skill can execute git worktree remove and git branch -D based on its analysis.
Sanitization: The script does not sanitize or escape commit subjects or filenames before they are included in the report provided to the AI agent.

worktree-tidy