issue-analyze

Issue Analyze Skill

Trigger

• Keywords: analyze issue, investigate problem, problem analysis, root cause, root cause analysis, triage review, review thread analysis

Input Types

| Type | Source | Example | |------|--------|---------| | GitHub Issue | Issue number, URL, or description | /issue-analyze 123 | | Review Thread | file:line + reviewer comment | /issue-analyze --triage "src/foo.ts:42 — Use early return" |

When input is a Review Thread:

• Phase 1 skips gh issue view, uses provided thread data directly
• Phase 2 classification uses Review Thread dimensions (see references/classification.md)

Modes

| Mode | Input Type | Phases Executed | Use Case | |------|-----------|----------------|----------| | Full (default) | Issue or Thread | 1 → 2 → 2.5 → 3 → 4 | Deep analysis with investigation | | Triage (--triage) | Review Thread | 2 → 2.5 only | Lightweight classification + verdict (thread data provided inline) | | Triage (--triage) | GitHub Issue | 1 → 2 → 2.5 | Fetch issue first, then classify + verdict |

When NOT to Use

• Known root cause, fix directly (use /bug-fix)
• Pure feature development (use /feature-dev)
• Only need code review (use /codex-review)

Workflow

┌─────────────────────────────────────────────────────────────────┐
│ Phase 1: Read Issue / Review Thread                             │
├─────────────────────────────────────────────────────────────────┤
│ GitHub Issue: gh issue view <number> --json ...                  │
│ Review Thread: use provided { path, line, reviewer, comment }   │
│ Extract: symptoms, reproduction steps, error messages, files    │
└─────────────────────────────────────────────────────────────────┘
                              ↓
┌─────────────────────────────────────────────────────────────────┐
│ Phase 2: Problem Classification                                 │
├─────────────────────────────────────────────────────────────────┤
│ GitHub Issue: Determine problem type → investigation strategy   │
│ Review Thread: Determine category → actionability               │
│                                                                 │
│ ┌────────────────┬──────────────────────────────────┐           │
│ │ Type           │ Investigation Strategy            │           │
│ ├────────────────┼──────────────────────────────────┤           │
│ │ Unfamiliar     │ /code-explore                    │           │
│ │ Regression     │ /git-investigate                 │           │
│ │ Complex root   │ /code-investigate (dual view)    │           │
│ │ Multiple cause │ /codex-brainstorm (exhaustive)   │           │
│ │ Composite      │ Combine multiple strategies      │           │
│ └────────────────┴──────────────────────────────────┘           │
└─────────────────────────────────────────────────────────────────┘
                              ↓
┌─────────────────────────────────────────────────────────────────┐
│ Phase 2.5: Verdict Assessment (NEW)                             │
├─────────────────────────────────────────────────────────────────┤
│ Codex blind verification: ACTIONABLE / NON_ACTIONABLE / UNCERTAIN│
│ Fresh mcp__codex__codex thread, read-only, anti-anchoring       │
│ Pattern: @skills/seek-verdict/references/verdict-prompt.md      │
│ Thresholds: @skills/seek-verdict/references/policy-mapping.md   │
│                                                                 │
│ ⚠️ Never send Claude's classification to Codex                  │
│ --triage mode: stop here, output verdict + classification       │
└─────────────────────────────────────────────────────────────────┘
                              ↓
┌─────────────────────────────────────────────────────────────────┐
│ Phase 3: Execute Investigation                                  │
├─────────────────────────────────────────────────────────────────┤
│ Skip if verdict = DISMISS_VERIFIED (per policy-mapping thresholds)    │
│ Otherwise: invoke corresponding investigation command           │
└─────────────────────────────────────────────────────────────────┘
                              ↓
┌─────────────────────────────────────────────────────────────────┐
│ Phase 4: Consolidated Report                                    │
├─────────────────────────────────────────────────────────────────┤
│ Synthesize investigation results + verdict, produce report      │
└─────────────────────────────────────────────────────────────────┘

Problem Classification Decision Tree

Issue Symptoms
    │
    ├─ "It used to work, now it doesn't" ───→ /git-investigate
    │                                          (find introduction point)
    │
    ├─ "Don't know how this feature works" ─→ /code-explore
    │                                          (quick understanding)
    │
    ├─ "Has error message / stack trace"
    │       │
    │       ├─ Clear error ────────────────→ /code-explore
    │       │                                (trace path)
    │       │
    │       └─ Vague / intermittent ───────→ /code-investigate
    │                                        (dual-view confirmation)
    │
    ├─ "Many possible causes" ────────────→ /codex-brainstorm
    │                                        (exhaustive analysis)
    │
    └─ Composite / uncertain ─────────────→ Start with /code-explore
                                             then choose based on results

Investigation Tool Comparison

| Tool | Purpose | Speed | Depth | | ------------------- | ---------------------- | ------- | ---------- | | /code-explore | Quick code exploration | Fast | Single | | /git-investigate | Track change history | Medium | Single | | /code-investigate | Dual confirmation | Slow | Dual-view | | /codex-brainstorm | Exhaust possibilities | Slowest | Adversarial|

Phase 2.5: Verdict Assessment

After classification, run Codex blind verification to independently assess actionability.

Codex call requirements (same as /seek-verdict pattern):

| Requirement | Detail | |-------------|--------| | Thread | Fresh mcp__codex__codex (never reuse existing thread) | | Sandbox | read-only | | Approval policy | never | | Anti-anchoring | Never send Claude's Phase 2 classification to Codex |

Prompt construction: follow @skills/seek-verdict/references/verdict-prompt.md pattern, adapting input:

• GitHub Issue input: provide issue title, body, labels as finding context
• Review Thread input: provide file path, line, reviewer comment as finding context
• Always include Standard Research Block (git status, git diff, grep, cat)

Codex output (all fields required):

- codex_verdict: ACTIONABLE | NON_ACTIONABLE | UNCERTAIN
- confidence: [0.0 - 1.0]
- evidence_refs: [files/lines/commands used]
- reasoning: [why this verdict]

Policy mapping: follow @skills/seek-verdict/references/policy-mapping.md thresholds (normal state; heightened thresholds apply after [DISMISS_PATTERN_WARN] — see policy-mapping.md Anti-Abuse Guard):

| Verdict | Confidence | Evidence Refs | Result | |---------|------------|---------------|--------| | NON_ACTIONABLE | >= 0.80 (normal) / >= 0.85 (heightened) | >= 2 (normal) / >= 3 (heightened) | Skip Phase 3 investigation (DISMISS_VERIFIED) | | ACTIONABLE | >= 0.70 | any | Proceed to Phase 3 (FIX_REQUIRED) | | UNCERTAIN / low | any | any | Proceed to Phase 3 (NEED_HUMAN) |

--triage mode: stop after Phase 2.5, output classification + verdict only.

Graceful degradation: if Codex call fails, log warning and proceed to Phase 3 without verdict.

Output

Full Report (default)

## Issue Analysis: <title>
- **Classification**: <problem type>
- **Verdict**: ACTIONABLE / NON_ACTIONABLE / UNCERTAIN (confidence: 0.XX)
- **Root cause hypothesis**: <analysis>
- **Investigation strategy**: <tools + plan>
- **Priority**: P0 / P1 / P2

Triage Report (--triage mode)

## Triage: