The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes data from external sources including requirement documents and backlog items. It includes an 'Untrusted Content Boundary' to manage these inputs.
Ingestion Points: Discovery workflows read artifacts and platform payloads as described in references/workflows.md.
Boundary Markers: Explicit instructions in SKILL.md define a boundary between untrusted fetched content and the agent context.
Capability Inventory: The skill utilizes platform-specific tools for item management and terminal commands for git analysis across multiple referenced files.
Sanitization: Six content sanitization guards are defined in SKILL.md to remove sensitive pathing, internal IDs, and potential secrets from outbound data.
[Command Execution]: Terminal-based git commands are used to analyze commit history and contributor patterns.
Evidence: Found in references/ado-pull-request.md for reviewer identification.
Context: These are standard development operations used for automation and are scoped to the local repository history.
[Credential Handling]: The skill instructions reference local environment files for authenticating with Jira.
Evidence: Mentions the use of ~/.jira.env in SKILL.md for the Jira command surface.
Context: This is a standard method for CLI tool authentication and is complemented by instructions to prevent these values from appearing in logs or outbound messages.

backlog-management