The Agent Skills Directory

[SAFE]: The skill incorporates a robust safety architecture through an 'Authority Boundary' mechanism documented in product-ownership-operating-model.md and vision.md. This architecture mandates that the agent escalate and receive human authorization for any action that is irreversible or outward-facing, such as public releases, commercial changes, or data deletion. This design ensures human oversight for high-impact activities.
[SAFE]: The skill implements a structured decision-making process using Product Decision Records (PDRs) defined in product-state-and-continuity.md. These records require the agent to document the context, options, and rationale for decisions, and include 'reversal triggers'—pre-committed conditions under which a decision should be revisited. This promotes transparency and prevents the agent from silently diverging from established product strategies.
[SAFE]: Integration with external tools, such as issue trackers and version control systems, is handled through a well-defined 'tracker adapter contract'. The instructions specifically advise the agent to verify tool commands against official documentation and local help files rather than relying on internal patterns, reducing the risk of command hallucination or misuse.

using-product-management