issue-tracking
Issue Tracking
Keep work resumable: what changed, what remains, who acts next, and what evidence supports the status. Reuse the project's tracker and conventions; no particular provider, spec workflow, or agent team is required.
Resolve the destination
- Choose in this order: the user's explicit destination for this work, the project's existing tracker/local record, then an explicitly configured personal or environment default only when no project destination exists. This skill supplies no provider default.
- Before falling back, check applicable project instructions, workflow docs, referenced private configuration, and the current task's primary record for an existing destination. Reuse context already loaded. If the mapping is incomplete, use relevant connected tools for targeted read-only discovery of the project/record; resolve ambiguous or conflicting mappings before writes, not by falling back. Incidental issue links are not a destination override.
- A GitHub remote, PR workflow, available
gh, or connected MCP does not select the issue tracker. GitHub Issues is eligible only through the same explicit choice, project convention, or configured fallback as any other provider. For example, a project managed in a separate tracker keeps its issues there even when its code and PRs live on GitHub. - An unavailable connector, missing permission, or failed lookup does not mean the project has no tracker. Preserve the destination and pending update, report it as not synced, and continue independent work; do not create a replacement issue in the default tracker.
- Reuse exact issue keys and known project IDs. Discover only missing identifiers and legal field values. For multiple repositories, subprojects, or trackers, identify the destination per work item and its primary record; do not mirror automatically.
- Keep provider, workspace/project, completion rules, and any local record location in existing project instructions or private configuration. No new config format is required. Keep private mappings and credentials out of this shared skill; preserve company/personal boundaries.
- If neither a project destination nor an applicable configured default can be established, ask once when tracking is needed; continue independent work. Reuse an agreed local task file when no external tracker is wanted. Preserve its format and stable item references rather than creating a second backlog.
Use the platform's model
Read the selected connector's tool descriptions and relevant usage guides before composing queries or writes; prefer its documented workflows. Inspect only metadata needed for the task. Current schemas and returned legal values own API mechanics; project conventions own how that platform is used. Do not copy tool signatures, status lists, or transient capability limits into this skill.
Use platform capabilities to improve retrieval, organization, and handoffs while preserving the agreed development workflow. Tool availability alone must not change development stages, acceptance criteria, assignment responsibilities, or checkpoint timing, or block otherwise authorized work; adapt the recording method and disclose any synchronization gap.