The Agent Skills Directory

[Integrated Safety Protocols]: The skill implements five mandatory safety layers, including human-review checkboxes and a three-tier autonomy model. These ensure that mutating operations, such as creating or updating work items, are subject to user-defined oversight and explicit destination confirmation.\n- [Untrusted Content Boundary]: The instructions explicitly designate item bodies and platform payloads as untrusted data. By treating these inputs as observed content rather than executable directives, the skill incorporates defenses against indirect prompt injection.\n- [Dry-Run Functionality]: A dedicated dry-run mode allows for full validation of operations without performing any actual mutations. This capability enables users to audit the exact data being sent to external trackers before execution.\n- [Credential Safety Measures]: The skill includes specific stop rules that trigger if a probable secret or credential is detected in the input, helping to prevent the accidental leakage of sensitive authentication information during processing.

backlog-execute