The Agent Skills Directory

[SAFE]: The skill functions as an advanced orchestration layer for development workflows. It emphasizes governance, agent boundaries, and quality control.
[COMMAND_EXECUTION]: The skill uses shell commands for repository metadata analysis (e.g., git log to detect change heatmaps and coupling) and local version control operations (e.g., git checkout for rollbacks). These operations are standard for development-oriented agents and are used for analysis rather than executing arbitrary remote code.
[PROMPT_INJECTION]: While the instructions contain strong directives like 'Hard rule', 'Non-Negotiable', and 'Iron Rule', these are used to enforce internal governance logic (such as ensuring the agent dispatches work instead of self-executing) and do not attempt to bypass platform safety guidelines.
[EXTERNAL_DOWNLOADS]: The skill references a specific ecosystem of 'Dependency Skills' (e.g., superpowers, agent-teams-playbook, gstack) that are expected to be installed via local project scripts (npm run deps:install). These references are documented as part of the framework's modular architecture.
[DATA_EXFILTRATION]: There is no evidence of sensitive data being sent to external or unauthorized domains. The network access mentioned (via browser) is for technical discovery and research consistent with agent design tasks.

meta-theory