The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to identify changed files using git diff and runs local repository linters such as eslint, prettier, and ruff based on detected configuration files.
[PROMPT_INJECTION]: The skill performs analysis on untrusted data from source code changes (Category 8 surface). It includes a specific instruction to treat processed content as data rather than instructions, mitigating the risk of indirect prompt injection from comments or strings. Ingestion points include git diff output in SKILL.md; boundary markers are explicitly provided via the 'data, not instructions' warning; capabilities include local linter execution and file writing; sanitization is performed via a mandatory user review table for all changes.
[DATA_EXFILTRATION]: The skill includes a mandatory safety check to identify and redact sensitive information like API keys or credentials found in the diff, ensuring they are not echoed in reports or review tables.

tanner-cleanup