Iterate Pr

1---
2name: iterate-pr
3description: Iterate on a PR until CI passes. Use when you need to fix CI failures, address review feedback, or continuously push fixes until all checks are green. Automates the feedback-fix-push-wait cycle.
4---
5 
6# Iterate on PR Until CI Passes
7 
8Continuously iterate on the current branch until all CI checks pass and review feedback is addressed.
9 
10**Requires**: GitHub CLI (`gh`) authenticated.
11 
12**Important**: All scripts must be run from the repository root directory (where `.git` is located), not from the skill directory. Use the full path to the script via `${CLAUDE_SKILL_ROOT}`.
13 
14## Bundled Scripts
15 
16### `scripts/fetch_pr_checks.py`
17 
18Fetches CI check status and extracts failure snippets from logs.
19 
20```bash
21uv run ${CLAUDE_SKILL_ROOT}/scripts/fetch_pr_checks.py [--pr NUMBER]
22```
23 
24Returns JSON:
25```json
26{
27  "pr": {"number": 123, "branch": "feat/foo"},
28  "summary": {"total": 5, "passed": 3, "failed": 2, "pending": 0},
29  "checks": [
30    {"name": "tests", "status": "fail", "log_snippet": "...", "run_id": 123},
31    {"name": "lint", "status": "pass"}
32  ]
33}
34```
35 
36### `scripts/fetch_pr_feedback.py`
37 
38Fetches and categorizes PR review feedback using the [LOGAF scale](https://develop.sentry.dev/engineering-practices/code-review/#logaf-scale).
39 
40```bash
41uv run ${CLAUDE_SKILL_ROOT}/scripts/fetch_pr_feedback.py [--pr NUMBER]
42```
43 
44Returns JSON with feedback categorized as:
45- `high` - Must address before merge (`h:`, blocker, changes requested)
46- `medium` - Should address (`m:`, standard feedback)
47- `low` - Optional (`l:`, nit, style, suggestion)
48- `bot` - Informational automated comments (Codecov, Dependabot, etc.)
49- `resolved` - Already resolved threads
50 
51Review bot feedback (from Sentry, Warden, Cursor, Bugbot, CodeQL, etc.) appears in `high`/`medium`/`low` with `review_bot: true` — it is NOT placed in the `bot` bucket.
52 
53## Workflow
54 
55### 1. Identify PR
56 
57```bash
58gh pr view --json number,url,headRefName
59```
60 
61Stop if no PR exists for the current branch.
62 
63### 2. Gather Review Feedback
64 
65Run `${CLAUDE_SKILL_ROOT}/scripts/fetch_pr_feedback.py` to get categorized feedback already posted on the PR.
66 
67### 3. Handle Feedback by LOGAF Priority
68 
69**Auto-fix (no prompt):**
70- `high` - must address (blockers, security, changes requested)
71- `medium` - should address (standard feedback)
72 
73This includes review bot feedback (items with `review_bot: true`). Treat it the same as human feedback:
74- Real issue found → fix it
75- False positive → skip, but explain why in a brief comment
76- Never silently ignore review bot feedback — always verify the finding
77 
78**Prompt user for selection:**
79- `low` - present numbered list and ask which to address:
80 
81```
82Found 3 low-priority suggestions:
831. [l] "Consider renaming this variable" - @reviewer in api.py:42
842. [nit] "Could use a list comprehension" - @reviewer in utils.py:18
853. [style] "Add a docstring" - @reviewer in models.py:55
86 
87Which would you like to address? (e.g., "1,3" or "all" or "none")
88```
89 
90**Skip silently:**
91- `resolved` threads
92- `bot` comments (informational only — Codecov, Dependabot, etc.)
93 
94### 4. Check CI Status
95 
96Run `${CLAUDE_SKILL_ROOT}/scripts/fetch_pr_checks.py` to get structured failure data.
97 
98**Wait if pending:** If review bot checks (sentry, warden, cursor, bugbot, seer, codeql) are still running, wait before proceeding—they post actionable feedback that must be evaluated. Informational bots (codecov) are not worth waiting for.
99 
100### 5. Fix CI Failures
101 
102For each failure in the script output:
1031. Read the `log_snippet` to understand the failure
1042. Read the relevant code before making changes
1053. Fix the issue with minimal, targeted changes
106 
107Do NOT assume what failed based on check name alone—always read the logs.
108 
109### 6. Commit and Push
110 
111```bash
112git add <files>
113git commit -m "fix: <descriptive message>"
114git push
115```
116 
117### 7. Wait for CI
118 
119```bash
120gh pr checks --watch --interval 30
121```
122 
123### 8. Re-check Feedback After CI
124 
125Review bots often post feedback seconds after CI checks complete. Wait briefly, then check again:
126 
127```bash
128sleep 10
129uv run ${CLAUDE_SKILL_ROOT}/scripts/fetch_pr_feedback.py
130```
131 
132Address any new high/medium feedback the same way as step 3. If new feedback requires code changes, return to step 6 to commit and push.
133 
134### 9. Repeat
135 
136Return to step 2 if CI failed or new feedback appeared in step 8.
137 
138## Exit Conditions
139 
140**Success:** All checks pass, post-CI feedback re-check is clean (no new unaddressed high/medium feedback including review bot findings), user has decided on low-priority items.
141 
142**Ask for help:** Same failure after 3 attempts, feedback needs clarification, infrastructure issues.
143 
144**Stop:** No PR exists, branch needs rebase.
145 
146## Fallback
147 
148If scripts fail, use `gh` CLI directly:
149- `gh pr checks name,state,bucket,link`
150- `gh run view <run-id> --log-failed`
151- `gh api repos/{owner}/{repo}/pulls/{number}/comments`
152