Skip to content

Commit 58c400f

Browse files
isPANNclaude
isPANN
and
claude
authored
Improve fix-issue and fix-issue-batch skills (#641)
- fix-issue: Add changelog comment step (8b), require Warn results to be fixed (not just Fail), simplify re-check to inline instead of subagent - fix-issue-batch: Dispatch each issue as a subagent for fresh context, present all questions/options as text for human to answer at once Addresses #636 (batch agent cutting corners). Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
1 parent deb5826 commit 58c400f

2 files changed

Lines changed: 517 additions & 0 deletions

File tree

‎.claude/skills/fix-issue-batch/SKILL.md‎

Lines changed: 242 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,242 @@
1+
---
2+
name: fix-issue-batch
3+
description: Batch-fix [Model] or [Rule] issues from the GitHub Project board Backlog column — calls /fix-issue on each one sequentially
4+
---
5+
6+
# Fix Issue Batch
7+
8+
Iterate through `[Model]` or `[Rule]` issues **from the Backlog column of the GitHub Project board** and call `/fix-issue` on each one sequentially.
9+
10+
## Invocation
11+
12+
```
13+
/fix-issue-batch <model|rule>
14+
```
15+
16+
## Constants
17+
18+
GitHub Project board IDs (same as fix-issue / project-pipeline):
19+
20+
| Constant | Value |
21+
|----------|-------|
22+
| `PROJECT_NUMBER` | `8` |
23+
| `PROJECT_OWNER` | `CodingThrust` |
24+
| `PROJECT_ID` | `PVT_kwDOBrtarc4BRNVy` |
25+
| `STATUS_FIELD_ID` | `PVTSSF_lADOBrtarc4BRNVyzg_GmQc` |
26+
| `STATUS_READY` | `61e4505c` |
27+
28+
## Process
29+
30+
```dot
31+
digraph fix_batch {
32+
rankdir=TB;
33+
"Parse argument" [shape=box];
34+
"Fetch project board Backlog" [shape=box];
35+
"Filter & sort" [shape=box];
36+
"Print queue" [shape=box];
37+
"Pick next issue" [shape=diamond];
38+
"Call /fix-issue N" [shape=box];
39+
"Done" [shape=doublecircle];
40+
41+
"Parse argument" -> "Fetch project board Backlog";
42+
"Fetch project board Backlog" -> "Filter & sort";
43+
"Filter & sort" -> "Print queue";
44+
"Print queue" -> "Pick next issue";
45+
"Pick next issue" -> "Call /fix-issue N" [label="remaining"];
46+
"Pick next issue" -> "Done" [label="none left"];
47+
"Call /fix-issue N" -> "Pick next issue";
48+
}
49+
```
50+
51+
---
52+
53+
## Step 1: Parse Argument
54+
55+
Accept one argument: `model` or `rule` (case-insensitive).
56+
57+
- `model` → filter for issues with `[Model]` in the title
58+
- `rule` → filter for issues with `[Rule]` in the title
59+
60+
If no argument or invalid argument → stop with: "Usage: `/fix-issue-batch <model|rule>`"
61+
62+
---
63+
64+
## Step 2: Fetch Issues from Project Board Backlog
65+
66+
Fetch all items from the GitHub Project board and filter to the **Backlog** column:
67+
68+
```bash
69+
gh project item-list 8 --owner CodingThrust --format json --limit 500
70+
```
71+
72+
From the JSON result:
73+
1. Filter items where `status == "Backlog"`
74+
2. Keep only items whose `content.title` starts with `[Model]` or `[Rule]` (matching the argument)
75+
3. For each item, extract: issue number (`content.number`), title (`content.title`), and project item ID (`id`)
76+
77+
Then fetch labels for each matching issue (needed for categorization):
78+
79+
```bash
80+
gh issue view <NUMBER> --repo CodingThrust/problem-reductions --json labels
81+
```
82+
83+
Or batch-fetch with a single search query to avoid N+1:
84+
85+
```bash
86+
gh issue list --repo CodingThrust/problem-reductions \
87+
--state open --search "[Model] in:title" \
88+
--json number,title,labels --limit 200
89+
```
90+
91+
Cross-reference the board items with the label data to build the final list.
92+
93+
---
94+
95+
## Step 3: Filter and Sort
96+
97+
From the Backlog issues:
98+
99+
1. **Categorize by current label status:**
100+
- `already-good` — has `Good` label
101+
- `has-failures` — has at least one failure label (`PoorWritten`, `Wrong`, `Trivial`, `Useless`)
102+
- `needs-check` — no check-issue comment yet (no `Good`/`PoorWritten`/`Wrong`/`Trivial`/`Useless` label)
103+
2. **Sort by priority then issue number:** `already-good` first, then `has-failures`, then `needs-check`. Within each group, sort by issue number ascending (oldest first). This prioritizes issues that already have a check report and are closest to being ready.
104+
105+
---
106+
107+
## Step 4: Print Queue
108+
109+
Show the user what will be processed:
110+
111+
```
112+
## Fix queue: [Model] issues (N total)
113+
114+
| # | Issue | Title | Status |
115+
|---|-------|-------|--------|
116+
| 1 | #235 | [Model] SteinerTree | already-good |
117+
| 2 | #233 | [Model] StrongConnectivityAugmentation | has-failures (PoorWritten) |
118+
| 3 | #234 | [Model] FeedbackVertexSet | needs-check |
119+
| ... | ... | ... | ... |
120+
121+
Priority: already-good → has-failures → needs-check.
122+
Processing N issues total.
123+
```
124+
125+
Then ask the user to confirm before starting:
126+
127+
> Ready to start? I'll process each issue with `/fix-issue`, one at a time.
128+
>
129+
> 1. **Start** — begin processing from the first issue
130+
> 2. **Start from #N** — skip to a specific issue number
131+
> 3. **Cancel**
132+
133+
---
134+
135+
## Step 5: Process Each Issue
136+
137+
**CRITICAL:** Each issue MUST be dispatched as a **subagent** for analysis, giving it a fresh context window to prevent cutting corners. The subagent does NOT interact with the human — it returns a structured report, then the main agent presents it for human decisions.
138+
139+
For each issue in the queue (priority order: `already-good` → `has-failures` → `needs-check`):
140+
141+
### 5a: Check prerequisite
142+
143+
If no comment starting with `## Issue Quality Check` exists, run `/check-issue <NUMBER>` first.
144+
145+
### 5b: Dispatch analysis subagent
146+
147+
```
148+
Agent tool:
149+
subagent_type: "general-purpose"
150+
description: "Analyze issue #<NUMBER>"
151+
prompt: |
152+
Analyze GitHub issue #<NUMBER> for fix-issue.
153+
1. Fetch the issue: gh issue view <NUMBER> --json title,body,labels,comments
154+
2. Find the most recent "## Issue Quality Check" comment
155+
3. Parse all Fail and Warn results (warnings are NOT ignorable)
156+
4. For each issue, classify as mechanical or substantive
157+
5. For mechanical issues: apply the fix to a draft body
158+
6. For substantive issues: prepare 2-3 concrete options with your recommendation
159+
160+
Return EXACTLY this format:
161+
162+
## Analysis for #<NUMBER>: <title>
163+
164+
### Auto-fixes applied
165+
| # | Section | Issue | Fix |
166+
|---|---------|-------|-----|
167+
| 1 | ... | ... | ... |
168+
169+
### Questions for human
170+
**Q1: <topic>**
171+
<description of the problem>
172+
- (a) <option 1> ← recommended
173+
- (b) <option 2>
174+
- (c) <option 3>
175+
176+
**Q2: ...**
177+
178+
### Draft body
179+
<full updated issue body with mechanical fixes applied, substantive issues marked as `[PENDING Q1]`>
180+
```
181+
182+
### 5c: Present to human
183+
184+
Print the subagent's report and ask the human to answer all questions at once:
185+
186+
```
187+
## Issue #<NUMBER> (<current>/<total>): <title>
188+
189+
<auto-fixes table from subagent>
190+
191+
<questions from subagent>
192+
193+
Please answer the questions above (e.g. "Q1: a, Q2: b"), or type "skip" to skip this issue.
194+
```
195+
196+
### 5d: Apply answers and finalize
197+
198+
After human responds:
199+
- If **"skip"**: move to next issue
200+
- Otherwise: apply the human's choices to the draft body, re-check (run 4 quality checks inline), then finalize on GitHub (edit body, post changelog comment, update labels, move to Ready — see fix-issue Steps 6–8)
201+
202+
### 5e: Continue
203+
204+
Print progress and ask whether to continue:
205+
206+
```
207+
Done #<NUMBER>. (<current>/<total> complete, <remaining> remaining)
208+
Next: #<NEXT_NUMBER> <next_title>
209+
```
210+
211+
> 1. **Continue** — process the next issue
212+
> 2. **Skip next** — skip the next issue and continue to the one after
213+
> 3. **Stop** — end the batch here and print summary
214+
215+
---
216+
217+
## Step 6: Summary
218+
219+
After all issues are processed, print a summary:
220+
221+
```
222+
## Batch fix complete
223+
224+
| Result | Count | Issues |
225+
|--------|-------|--------|
226+
| Fixed & moved to Ready | 7 | #233, #234, #235, #236, #237, #238, #240 |
227+
| Skipped (by user) | 1 | #239 |
228+
| Total | 8 | |
229+
```
230+
231+
---
232+
233+
## Common Mistakes
234+
235+
| Mistake | Fix |
236+
|---------|-----|
237+
| Fetching all open issues instead of board | Only process issues from the **Backlog** column of GitHub Project #8 |
238+
| Skipping already-good issues | Process ALL Backlog issues; `already-good` are processed first (highest priority) |
239+
| Not running check-issue first | If no check report exists, run `/check-issue` before `/fix-issue` |
240+
| Processing in random order | Always sort by issue number ascending within priority groups |
241+
| Continuing after user cancels | Respect cancel/skip requests immediately |
242+
| Missing project scopes | Run `gh auth refresh -s read:project,project` if board access fails |

0 commit comments

Comments
 (0)