The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script "scripts/run_quality_gate.py" dynamically modifies the Python path using "Path(file).parent" to load the "quality_gate" module from a computed relative location. While typical for shared skill libraries, this represents dynamic loading from a computed path.
[INDIRECT_PROMPT_INJECTION]: The skill scans repository files for technical debt, including TODOs and FIXMEs, which are untrusted data that could contain malicious instructions designed to influence agent behavior. (1) Ingestion points: "Read", "Grep", and "Glob" tools are used to ingest repository code. (2) Boundary markers: No explicit delimiters or instructions are used to distinguish scanned code from system instructions. (3) Capability inventory: The skill has "Bash", "Write", "Edit", and "Task" capabilities. (4) Sanitization: No sanitization is applied to the content extracted from comments.
[EXTERNAL_DOWNLOADS]: Reference documentation (e.g., "references/code-cleanup.md") suggests installing third-party tools such as "ruff", "vulture", and "gocyclo" using standard package managers from public repositories. The use of "npx" in linting commands also implies potential external package retrieval.

code-quality