The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands including find, grep, awk, wc, and sort to perform pattern matching and file analysis within the turbo/ directory. These are standard operations for a static analysis tool.
[COMMAND_EXECUTION]: Uses the official GitHub CLI (gh) to programmatically create issues and comments. This involves sending analyzed code snippets and file paths to GitHub, which is a well-known service and the intended destination for the skill's output.
[DATA_EXFILTRATION]: While the skill extracts source code snippets and project metadata to populate GitHub issues, this behavior is documented as the primary function of the skill for technical debt tracking. No evidence of unauthorized data transfer to untrusted domains was found.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted source code within the repository (Phase 2). Although there are no specific boundary markers or sanitization steps mentioned for the ingested code, the capabilities are limited to reporting, which poses a low risk of systemic compromise.
Ingestion points: Files within the turbo/ directory are read in full during the detailed analysis phase.
Boundary markers: None identified in the prompt templates for report generation.
Capability inventory: Shell execution (filesystem utilities) and GitHub CLI operations (gh issue create, gh issue comment).
Sanitization: No explicit sanitization or escaping of the ingested code snippets is described before they are included in the reports.

tech-debt