name: agent-ops-tasks description: "Create, refine, and manage issues. Use for creating new issues from loose ideas, refining ambiguous issues, bulk operations, or JSON export." category: core invokes: [agent-ops-state, agent-ops-interview] invoked_by: [agent-ops-planning, agent-ops-focus-scan, agent-ops-report] state_files: read: [focus.md, issues/.md] write: [focus.md, issues/.md] reference: [REFERENCE.md]
Issue Management
Works with or without aoc CLI installed. All operations can be performed via direct file editing.
CRITICAL: Issue Management ONLY
This skill manages issues. It NEVER implements code.
- ✅ Create, refine, list, search, triage issues
- ✅ Move issues between priority files
- ❌ NEVER implement features or fix bugs
- ❌ NEVER modify code files
After any issue operation, ALWAYS offer a handoff — never auto-proceed.
Reference: See REFERENCE.md for templates, CLI commands, JSON export.
Issue ID Format
Format: {TYPE}-{NUMBER}@{HASH}
Example: BUG-0023@efa54f, FEAT-0001@c2d4e6
Types: BUG | FEAT | CHORE | ENH | SEC | PERF | DOCS | TEST | REFAC | PLAN
Minimal Issue Template
## {TYPE}-{NUMBER}@{HASH} — {title}
id: {TYPE}-{NUMBER}@{HASH}
title: "{title}"
type: {type}
status: todo | in_progress | done
priority: critical | high | medium | low
description: {brief description}
details: references/{TYPE}-{NUMBER}@{HASH}.md
### Acceptance Criteria
- [ ] Criterion 1
### Log
- YYYY-MM-DD: Created
Issue Size Guardrails
- Keep backlog items minimal: title, metadata, 1–2 sentence description, acceptance criteria if known.
- If an issue needs more than ~20 lines, move details to a reference file in
.agent/issues/references/and link it in the issue. - Reference files should contain research, long descriptions, examples, diagrams, or interview notes.
- Never embed large code blocks or research dumps directly in backlog items.
Reference File Format
- Path:
.agent/issues/references/{ISSUE-ID}.md - Include a short header and a link back to the issue.
- Example:
# {ISSUE-ID} — {title}
Moved from backlog.md on YYYY-MM-DD.
## Context
...
File Organization
| File | Priority |
|---|---|
.agent/issues/critical.md | Blockers, production issues |
.agent/issues/high.md | Important, address soon |
.agent/issues/medium.md | Standard work |
.agent/issues/low.md | Nice-to-have |
.agent/issues/backlog.md | Unprioritized ideas |
.agent/issues/history.md | Completed/archived |
Operations
Create Issue
- Analyze request for type, title, priority, scope, criteria
- Use
agent-ops-interviewfor missing info - Generate ID from
.agent/issues/.counter - Create issue, append to priority file
- STOP AND HANDOFF
Mandatory Handoff
✅ Issue created: {ISSUE-ID}: {title}
What's next?
1. Start implementing (requires confirmation)
2. Create more issues
3. Do nothing
Refine Issue
Triggers for refinement:
- Generic titles ("Fix bugs")
- Missing acceptance criteria
- Confidence marked
low
Procedure: Interview for scope, criteria, dependencies, risks.
Change Priority
- Remove from current file
- Update
priorityfield - Add log entry
- Append to new file
Triage Backlog
For each backlog item: assign priority or skip/delete.
Issue Discovery
Other skills invoke discovery when they find potential work:
| Skill | Triggers |
|---|---|
baseline | Warnings, failures, missing coverage |
planning | Sub-tasks, prerequisites |
critical-review | Bugs, security, tech debt |
Procedure:
- Collect findings
- Categorize by type/priority
- Present summary to user
- Create issues on confirmation
- Offer next actions
Quality Checklist
- Valid ID format
- Action-oriented title
- Testable acceptance criteria
- Appropriate confidence level