The Agent Skills Directory

[SAFE]: The skill implements a multi-stage validation process that includes orientations, structural checks, spike executions, and a final polishing phase. This structured approach, combined with a mandatory human approval gate, ensures high oversight.
[COMMAND_EXECUTION]: The skill uses local command-line tools br (beads-cli) and bv (beads-viewer) to manage project tasks and visualize dependency graphs. These operations are constrained to the local development environment and are used to verify the integrity of project plans.
[PROMPT_INJECTION]: Instructions for sub-agents ('plan-checker' and 'bead-reviewer') are provided to perform specific analytical tasks. These prompts establish clear roles and constraints to prevent the sub-agents from diverging from the validation objective.
[PROMPT_INJECTION]: The skill processes project-related markdown files and task data which may contain external or user-provided content. This ingestion surface is managed through the use of isolated sub-agents and final manual review. The analysis identified the following surface characteristics: ingestion points include history and state files; boundary markers between instructions and data are not explicitly defined in the sub-agent prompts; capabilities include state modification via CLI tools; and no explicit sanitization of the input data is performed.

validating