The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes multiple Bash scripts (generate-blind-visualization.sh, generate-exhaustive-spec.sh, run-adversarial-refinement.sh, validate-spec-completeness.sh) intended for local use. These scripts manage file system operations such as directory creation and markdown template generation. The generate-exhaustive-spec.sh script specifically implements security checks to prevent path traversal by ensuring output directories do not use relative parent paths (..) or absolute root paths.
[SAFE]: No malicious patterns such as prompt injection, credential harvesting, or data exfiltration were detected. The methodology focuses on 'adversarial thinking' as a positive quality-assurance measure for documentation accuracy rather than a means to bypass AI safety guardrails.
[DATA_EXFILTRATION]: The skill's logic is self-contained and operates exclusively on local project files. There are no network-capable commands (e.g., curl, wget) or logic that attempts to access sensitive system directories like .ssh or .aws.
[EXTERNAL_DOWNLOADS]: The scripts do not download external dependencies or execute remote code. All operations are performed using standard local shell utilities.

exhaustive-specification