fix(bugfix): comprehensive workflow improvements addressing 9 open issues#90
fix(bugfix): comprehensive workflow improvements addressing 9 open issues#90jwm4 wants to merge 9 commits intoambient-code:mainfrom
Conversation
…sues Controller (ambient-code#81): Require AskUserQuestion tool for phase gates instead of prose "stop and wait" instructions. Add rules enforcing skill file reads and preventing urgency from bypassing process. Assess (ambient-code#82): Add "Check for Existing Work" step that scans open PRs and related issues before deep analysis. Add "not actually a bug" assessment path so the agent doesn't proceed to fix something it concluded isn't broken. Reproduce (ambient-code#57): Add environment setup recipes — detect project type from config files, provide setup commands per package manager, check for version managers, and include error recovery patterns for common failures. Fix (ambient-code#58): Add error handling UX review step. Guide agent to match error types to their context (CLI vs config vs runtime), test error paths manually, and ensure messages are clear and actionable. Test (ambient-code#55, ambient-code#56, ambient-code#59): Add "Survey Existing Test Patterns" step requiring the agent to examine existing tests before writing new ones. Make full test suite run mandatory with explicit "do not skip" language. Add dedicated linting/formatting step after writing test code. PR (ambient-code#80, ambient-code#85): Replace hardcoded "main" with DEFAULT_BRANCH placeholder detected from the upstream repo. Replace hardcoded "fork" remote name with FORK_REMOTE placeholder set during remote configuration. Update all downstream references (push, PR create, compare URLs, sync commands). Addresses ambient-code#55, ambient-code#56, ambient-code#57, ambient-code#58, ambient-code#59, ambient-code#80, ambient-code#81, ambient-code#82, ambient-code#85 Made-with: Cursor
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
WalkthroughAdded an assess gate that queries GitHub for existing PRs and pauses via AskUserQuestion on likely duplicates; mandated reading phase SKILL.md before executing phases and replaced ad-hoc waits with AskUserQuestion gates; added a dispatchable final Changes
Sequence Diagram(s)sequenceDiagram
participant Controller
participant Assess as AssessSkill
participant GitHub
participant User
Note over Controller,Assess: Assess phase runs (must Read SKILL.md first)
Controller->>Assess: Invoke Assess SKILL
Assess->>GitHub: Query issue-linked PRs & list open PRs
GitHub-->>Assess: Return PRs & titles
alt Directly addressing open PR found
Assess->>Controller: Pause & AskUserQuestion (present duplicates/related PRs)
Controller->>User: Notify & present choices (stop / continue / close / mark duplicate)
User-->>Controller: Selection
Controller->>Assess: Resume or stop based on response
else No related work found
Assess->>Controller: Continue to next step
end
sequenceDiagram
participant Controller
participant Summary as SummarySkill
participant Artifacts
participant User
Note over Controller,Summary: /summary can be invoked at any point
Controller->>Summary: Invoke /summary phase
Summary->>Artifacts: List artifacts/bugfix/*.md (exclude summary.md)
alt No artifacts
Summary->>Controller: Report "no artifacts" and stop
else Artifacts found
Summary->>Artifacts: Read each artifact
Summary->>Summary: Extract & categorize findings (existing work, understanding, root cause, tests, PR status, outstanding items)
Summary->>Artifacts: Write artifacts/bugfix/summary.md
Summary->>Controller: Return summary file path
Controller->>User: Present consolidated Bugfix Workflow Summary
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes 🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
When the assess skill finds an open PR that appears to address the bug, stop and use AskUserQuestion before continuing — even in speedrun mode. This prevents duplicate work when a fix is already in flight. The gate is in the skill rather than the controller/speedrun because "should we even continue this workflow?" is a discovery-time decision, and a tool call is structurally reliable in both interactive and speedrun modes (unlike prose escalation framing). Follow-up to the comprehensive update in the previous commit. Addresses ambient-code#82 Made-with: Cursor
New skill that scans artifacts/bugfix/ dynamically and presents a synthesized summary of findings, decisions, and status. Designed to surface information that gets buried in earlier artifacts — related PRs, review concerns, unconfirmed assumptions, outstanding items. Key design choices: - Discovers artifacts by scanning the directory tree rather than maintaining a hardcoded list (won't break when other skills change) - Invocable at any point mid-workflow, not just at the end - Added as phase 9 in both controller and speedrun flows - In speedrun, replaces the static completion report with a content-aware summary — the last thing the user sees Updated controller, speedrun, and CLAUDE.md to include the new phase. Made-with: Cursor
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
workflows/bugfix/.claude/skills/assess/SKILL.md (1)
197-203:⚠️ Potential issue | 🟡 MinorStep reference is stale after renumbering.
Line 202 says the artifact should mirror content from Step 6, but presentation is now in Step 7. This can cause process drift when operators copy the wrong section.
Proposed fix
-in Step 6. +in Step 7.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@workflows/bugfix/.claude/skills/assess/SKILL.md` around lines 197 - 203, Update the Step 8 description so it references the correct preceding step (Step 7) instead of the stale "Step 6"; specifically, edit the text that instructs saving the assessment to artifacts/bugfix/reports/assessment.md to state the file should contain the same content presented in Step 7 (the "presentation" step) and verify the artifact path and filename remain unchanged. Ensure the sentence mentioning "Step 6" is replaced with "Step 7" and that the wording clearly ties the artifact content to the presentation in Step 7 to avoid operator confusion.workflows/bugfix/.claude/skills/speedrun/SKILL.md (1)
26-27:⚠️ Potential issue | 🟠 MajorConflicting end-of-flow behavior after adding
/summary.Line 26 still mandates printing the Completion Report when all phases are done, but Line 134 says
/summary“replaces” that report. This creates contradictory terminal behavior and can produce duplicate end outputs.Proposed doc fix
-2. If all phases are done, print the completion report and stop +2. If all phases are done, stop (the `/summary` phase is the terminal user-facing report)-## Completion Report - -When all phases are done (or if you stop early due to escalation), present: +## Completion Report (Escalation/Early Stop Only) + +If you stop early due to escalation (before `/summary`), present:Also applies to: 132-157
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@workflows/bugfix/.claude/skills/speedrun/SKILL.md` around lines 26 - 27, The document has conflicting terminal behavior: line referencing "print the completion report and stop" conflicts with the note that the `/summary` command "replaces" that report; update SKILL.md so the end-of-flow behavior is consistent by specifying that when the `/summary` flag is present the normal "Completion Report" is suppressed and replaced by the `/summary` output, and when `/summary` is absent the Completion Report is printed as before; apply this clarification to the earlier flow description (the "If all phases are done" step) and the later "/summary" section (around lines 132-157) so both places state the single consistent rule.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@workflows/bugfix/.claude/skills/pr/SKILL.md`:
- Around line 362-365: The git diff invocation used to set WORKFLOW_DIFF places
the --name-only option after the -- pathspec separator, so Git treats it as a
path instead of an option; update the WORKFLOW_DIFF assignment so the
--name-only flag appears before the -- delimiter in the git diff command (i.e.,
call git diff FORK_REMOTE/DEFAULT_BRANCH..DEFAULT_BRANCH --name-only --
.github/workflows/), ensuring WORKFLOW_DIFF contains only filenames; locate the
assignment to WORKFLOW_DIFF in the file and reorder the arguments accordingly.
In `@workflows/bugfix/.claude/skills/summary/SKILL.md`:
- Around line 35-36: The discovery command currently lists all .md files ("find
artifacts/bugfix/ -type f -name '*.md' ... | sort") which will re-include the
generated summary (summary.md); update that command to explicitly exclude
summary.md (e.g., add -not -name 'summary.md' or an equivalent --exclude filter)
so the generated file is never read back in, and ensure any code that writes the
summary (the block that outputs summary.md) remains unchanged but is referenced
as the produced artifact to avoid self-ingestion.
In `@workflows/bugfix/.claude/skills/test/SKILL.md`:
- Around line 129-137: Replace the placeholder "FILE1 FILE2 ..." comments after
"git diff --name-only HEAD" with concrete examples showing how to pipe or
substitute that command into each formatter; update the Python, Node.js and Go
comment lines (the comments mentioning black/isort/ruff, npx prettier/eslint,
and gofmt) to show either command substitution using "$(git diff --name-only
HEAD | grep '\.py$')" / similar for js/ts/go or an xargs form "git diff
--name-only HEAD | grep '\.py$' | xargs black ..." so readers can see exactly
how to feed the git diff output into black/isort/ruff, prettier/eslint and
gofmt.
- Around line 36-41: The shell glob "cat *_test.go 2>/dev/null | head -5" can
fail when the glob doesn't expand; replace that pipeline with a find-based
search that explicitly looks for files matching "*_test.go" (limit search
depth), runs head on each matched file, redirects find/head stderr to /dev/null,
and then limits the overall output (e.g., to 20 lines); update the line
referencing "cat *_test.go" in SKILL.md to use this find+head approach so
missing matches are handled robustly.
- Around line 75-78: The guidance currently includes a Python-specific example
using call_args.args[0] vs call_args[0][0] without qualification; update
SKILL.md to either (a) prepend a clear language qualifier (e.g., "Python
(unittest.mock):") when mentioning call_args.args[0 and call_args[0][0]) or (b)
move that example into the "Python projects" section so cross-language guidance
remains generic; edit the paragraph containing call_args.args[0 and
call_args[0][0] to add the qualifier or relocate the sentence to the Python
projects section and ensure the surrounding generic recommendations stay
language-agnostic.
---
Outside diff comments:
In `@workflows/bugfix/.claude/skills/assess/SKILL.md`:
- Around line 197-203: Update the Step 8 description so it references the
correct preceding step (Step 7) instead of the stale "Step 6"; specifically,
edit the text that instructs saving the assessment to
artifacts/bugfix/reports/assessment.md to state the file should contain the same
content presented in Step 7 (the "presentation" step) and verify the artifact
path and filename remain unchanged. Ensure the sentence mentioning "Step 6" is
replaced with "Step 7" and that the wording clearly ties the artifact content to
the presentation in Step 7 to avoid operator confusion.
In `@workflows/bugfix/.claude/skills/speedrun/SKILL.md`:
- Around line 26-27: The document has conflicting terminal behavior: line
referencing "print the completion report and stop" conflicts with the note that
the `/summary` command "replaces" that report; update SKILL.md so the
end-of-flow behavior is consistent by specifying that when the `/summary` flag
is present the normal "Completion Report" is suppressed and replaced by the
`/summary` output, and when `/summary` is absent the Completion Report is
printed as before; apply this clarification to the earlier flow description (the
"If all phases are done" step) and the later "/summary" section (around lines
132-157) so both places state the single consistent rule.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 6ed28089-79fe-42b9-8e8c-d04aaff674ad
📒 Files selected for processing (9)
workflows/bugfix/.claude/skills/assess/SKILL.mdworkflows/bugfix/.claude/skills/controller/SKILL.mdworkflows/bugfix/.claude/skills/fix/SKILL.mdworkflows/bugfix/.claude/skills/pr/SKILL.mdworkflows/bugfix/.claude/skills/reproduce/SKILL.mdworkflows/bugfix/.claude/skills/speedrun/SKILL.mdworkflows/bugfix/.claude/skills/summary/SKILL.mdworkflows/bugfix/.claude/skills/test/SKILL.mdworkflows/bugfix/CLAUDE.md
| ```bash | ||
| # Check for test configuration | ||
| cat pytest.ini 2>/dev/null || cat setup.cfg 2>/dev/null | head -20 | ||
| cat jest.config.* 2>/dev/null || cat vitest.config.* 2>/dev/null | ||
| cat *_test.go 2>/dev/null | head -5 | ||
| ``` |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Improve shell glob robustness for Go test file detection.
The command cat *_test.go 2>/dev/null | head -5 may behave unexpectedly if no matching files exist. In shells without nullglob enabled, the glob pattern *_test.go won't expand and cat will attempt to read a file literally named *_test.go, which will fail (silently due to 2>/dev/null, but still incorrect).
🔧 Proposed fix using a more robust file detection pattern
```bash
# Check for test configuration
cat pytest.ini 2>/dev/null || cat setup.cfg 2>/dev/null | head -20
cat jest.config.* 2>/dev/null || cat vitest.config.* 2>/dev/null
-cat *_test.go 2>/dev/null | head -5
+find . -maxdepth 2 -name '*_test.go' -exec head -5 {} \; 2>/dev/null | head -20
This uses `find` to explicitly search for matching files, avoiding glob expansion issues.
</details>
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against the current code and only fix it if needed.
In @workflows/bugfix/.claude/skills/test/SKILL.md around lines 36 - 41, The
shell glob "cat _test.go 2>/dev/null | head -5" can fail when the glob doesn't
expand; replace that pipeline with a find-based search that explicitly looks for
files matching "_test.go" (limit search depth), runs head on each matched file,
redirects find/head stderr to /dev/null, and then limits the overall output
(e.g., to 20 lines); update the line referencing "cat *_test.go" in SKILL.md to
use this find+head approach so missing matches are handled robustly.
</details>
<!-- fingerprinting:phantom:triton:puma -->
<!-- This is an auto-generated comment by CodeRabbit -->
| ```bash | ||
| # Identify all modified files | ||
| git diff --name-only HEAD | ||
|
|
||
| # Then run the appropriate formatters on ALL of them: | ||
| # Python: black FILE1 FILE2 ... && isort FILE1 FILE2 ... && ruff check --fix FILE1 FILE2 ... | ||
| # Node.js: npx prettier --write FILE1 FILE2 ... && npx eslint --fix FILE1 FILE2 ... | ||
| # Go: gofmt -w FILE1 FILE2 ... | ||
| ``` |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Consider adding a concrete example for applying formatters to modified files.
The placeholder syntax FILE1 FILE2 ... in lines 134-136 may not be immediately clear on how to integrate with the git diff --name-only HEAD output. Consider adding a concrete example using command substitution or xargs to show the connection between the two steps.
📋 Example using command substitution
```bash
# Identify all modified files
git diff --name-only HEAD
-# Then run the appropriate formatters on ALL of them:
-# Python: black FILE1 FILE2 ... && isort FILE1 FILE2 ... && ruff check --fix FILE1 FILE2 ...
-# Node.js: npx prettier --write FILE1 FILE2 ... && npx eslint --fix FILE1 FILE2 ...
-# Go: gofmt -w FILE1 FILE2 ...
+# Then run the appropriate formatters on ALL of them (example with command substitution):
+# Python: black $(git diff --name-only HEAD | grep '\.py$') && isort $(git diff --name-only HEAD | grep '\.py$')
+# Node.js: npx prettier --write $(git diff --name-only HEAD | grep '\.(js|ts|tsx)$')
+# Go: gofmt -w $(git diff --name-only HEAD | grep '\.go$')
Alternatively, mention `xargs` as another approach for handling file lists.
</details>
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against the current code and only fix it if needed.
In @workflows/bugfix/.claude/skills/test/SKILL.md around lines 129 - 137,
Replace the placeholder "FILE1 FILE2 ..." comments after "git diff --name-only
HEAD" with concrete examples showing how to pipe or substitute that command into
each formatter; update the Python, Node.js and Go comment lines (the comments
mentioning black/isort/ruff, npx prettier/eslint, and gofmt) to show either
command substitution using "$(git diff --name-only HEAD | grep '.py$')" /
similar for js/ts/go or an xargs form "git diff --name-only HEAD | grep '.py$'
| xargs black ..." so readers can see exactly how to feed the git diff output
into black/isort/ruff, prettier/eslint and gofmt.
</details>
<!-- fingerprinting:phantom:triton:puma -->
<!-- This is an auto-generated comment by CodeRabbit -->
There was a problem hiding this comment.
Actionable comments posted: 5
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
workflows/bugfix/.claude/skills/assess/SKILL.md (1)
202-202:⚠️ Potential issue | 🟡 MinorFix the step reference after renumbering.
After adding the new Step 3, "Present to the User" is now Step 7, not Step 6. The assessment artifact should contain the content from Step 7.
📝 Proposed fix
-in Step 6. +in Step 7.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@workflows/bugfix/.claude/skills/assess/SKILL.md` at line 202, The document still references "Step 6" for the assessment artifact but after adding a new Step 3 the "Present to the User" step is now Step 7; update the SKILL.md content that mentions "Step 6" (the assessment artifact reference) to point to Step 7 and ensure the assessment artifact includes the content from "Present to the User" (the updated step title) so the artifact pulls the correct step content.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@workflows/bugfix/.claude/skills/assess/SKILL.md`:
- Line 84: The hardcoded "--limit 30" in the gh pr list and corresponding issue
listing command should be justified or made configurable; update the SKILL.md
around the gh pr list --limit 30 invocation (and the duplicate at the other
occurrence) to either add a short comment explaining the trade-off (fast listing
vs missing PRs in high-activity repos) or document how to change the limit
(e.g., run the command with a different --limit or remove it), and ensure both
occurrences mention the rationale and how to adjust the limit so readers know
why 30 was chosen and how to increase it when needed.
- Line 78: The existing grep pattern "pull\|PR\|#" is too broad; update the grep
invocation in the pipeline command to use a stricter regex (and -E for extended
regex) that only matches explicit PR/issue references, e.g. replace the pattern
with something like
'(^|[^A-Za-z0-9])(#\d+|PR[[:space:]]*#\d+|pull[[:space:]]+request[[:space:]]*#\d+)([^A-Za-z0-9]|$)'
so the command that contains the current "pull\|PR\|#" pattern only returns real
references like "#123", "PR `#123`" or "pull request `#123`".
- Around line 96-106: The pr phase uses plain text "Stop again and wait for the
user to confirm" instead of invoking the AskUserQuestion tool; update the pr
phase SKILL.md to call AskUserQuestion with the same three options used in the
assess skill ("PR `#N` appears to address this bug — review it instead of starting
fresh", "PR `#N` is related but doesn't fully cover it — continue with
assessment", "Not sure if PR `#N` is relevant — continue with assessment") and
enforce the hard gate so the flow does not proceed until AskUserQuestion returns
the user's choice.
In `@workflows/bugfix/.claude/skills/pr/SKILL.md`:
- Around line 362-365: The git diff invocation assigned to WORKFLOW_DIFF places
the --name-only option after the pathspec which violates git syntax; update the
git diff call inside the WORKFLOW_DIFF assignment to pass --name-only as an
option before the revision range (FORK_REMOTE/DEFAULT_BRANCH..DEFAULT_BRANCH)
and keep the pathspec separator (--) before .github/workflows/ so the command
returns only file names reliably.
In `@workflows/bugfix/.claude/skills/speedrun/SKILL.md`:
- Around line 132-137: Update the "summary" skill description to explicitly
preserve mandatory interactive gates by name: state that the summary phase (the
"summary" section and the Speedrun behavior) must NOT bypass any hard
AskUserQuestion gates (e.g., "assess existing-PR decision") and must pause and
surface those gates for user input before continuing; reference the "summary"
skill and the AskUserQuestion gate type so readers know this final-phase
behavior still enforces earlier interactive checkpoints.
---
Outside diff comments:
In `@workflows/bugfix/.claude/skills/assess/SKILL.md`:
- Line 202: The document still references "Step 6" for the assessment artifact
but after adding a new Step 3 the "Present to the User" step is now Step 7;
update the SKILL.md content that mentions "Step 6" (the assessment artifact
reference) to point to Step 7 and ensure the assessment artifact includes the
content from "Present to the User" (the updated step title) so the artifact
pulls the correct step content.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 17de1b89-1db5-40c1-91d4-31abd08abc3a
📒 Files selected for processing (9)
workflows/bugfix/.claude/skills/assess/SKILL.mdworkflows/bugfix/.claude/skills/controller/SKILL.mdworkflows/bugfix/.claude/skills/fix/SKILL.mdworkflows/bugfix/.claude/skills/pr/SKILL.mdworkflows/bugfix/.claude/skills/reproduce/SKILL.mdworkflows/bugfix/.claude/skills/speedrun/SKILL.mdworkflows/bugfix/.claude/skills/summary/SKILL.mdworkflows/bugfix/.claude/skills/test/SKILL.mdworkflows/bugfix/CLAUDE.md
| - **Check for linked PRs on the issue:** | ||
|
|
||
| ```bash | ||
| gh issue view NUMBER --repo OWNER/REPO --json body,comments --jq '[.body, .comments[].body] | join("\n")' | grep -i "pull\|PR\|#" |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Refine the grep pattern to reduce false positives.
The pattern "pull\|PR\|#" is very broad and will match:
- Any markdown heading (
#) - Partial word matches (
"PR"in "PRIORITIZE") - Generic uses of "pull" unrelated to PRs
🔍 Suggested pattern refinement
-gh issue view NUMBER --repo OWNER/REPO --json body,comments --jq '[.body, .comments[].body] | join("\n")' | grep -i "pull\|PR\|#"
+gh issue view NUMBER --repo OWNER/REPO --json body,comments --jq '[.body, .comments[].body] | join("\n")' | grep -iE '#[0-9]+|PR #[0-9]+|pull request #[0-9]+'This matches actual PR/issue references like #123, PR #123, or `pull request `#123 rather than arbitrary occurrences.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| gh issue view NUMBER --repo OWNER/REPO --json body,comments --jq '[.body, .comments[].body] | join("\n")' | grep -i "pull\|PR\|#" | |
| gh issue view NUMBER --repo OWNER/REPO --json body,comments --jq '[.body, .comments[].body] | join("\n")' | grep -iE '#[0-9]+|PR #[0-9]+|pull request #[0-9]+' |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@workflows/bugfix/.claude/skills/assess/SKILL.md` at line 78, The existing
grep pattern "pull\|PR\|#" is too broad; update the grep invocation in the
pipeline command to use a stricter regex (and -E for extended regex) that only
matches explicit PR/issue references, e.g. replace the pattern with something
like
'(^|[^A-Za-z0-9])(#\d+|PR[[:space:]]*#\d+|pull[[:space:]]+request[[:space:]]*#\d+)([^A-Za-z0-9]|$)'
so the command that contains the current "pull\|PR\|#" pattern only returns real
references like "#123", "PR `#123`" or "pull request `#123`".
| - **Scan recent PR titles for overlap:** | ||
|
|
||
| ```bash | ||
| gh pr list --repo OWNER/REPO --state open --limit 30 --json number,title,headRefName --jq '.[] | "\(.number)\t\(.title)"' |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Consider documenting the rationale for --limit 30.
The hardcoded limit of 30 for both PR and issue listings is reasonable for most repositories, but relevant work could be missed in high-activity projects. Consider adding a comment explaining this trade-off, or note that users can manually check if they suspect additional relevant work exists.
Also applies to: 93-93
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@workflows/bugfix/.claude/skills/assess/SKILL.md` at line 84, The hardcoded
"--limit 30" in the gh pr list and corresponding issue listing command should be
justified or made configurable; update the SKILL.md around the gh pr list
--limit 30 invocation (and the duplicate at the other occurrence) to either add
a short comment explaining the trade-off (fast listing vs missing PRs in
high-activity repos) or document how to change the limit (e.g., run the command
with a different --limit or remove it), and ensure both occurrences mention the
rationale and how to adjust the limit so readers know why 30 was chosen and how
to increase it when needed.
| If you find an open PR that appears to directly address this bug, **stop here | ||
| and use `AskUserQuestion`** before continuing the assessment. Present the | ||
| options: | ||
|
|
||
| - "PR #N appears to address this bug — review it instead of starting fresh" | ||
| - "PR #N is related but doesn't fully cover it — continue with assessment" | ||
| - "Not sure if PR #N is relevant — continue with assessment" | ||
|
|
||
| This gate applies in both normal and speedrun mode. Do not continue to Step 4 | ||
| until the user responds. The `AskUserQuestion` tool triggers platform | ||
| notifications so the user knows you need their input. |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Description: Check for plain-text wait instructions instead of AskUserQuestion in skill files.
# Search for wait/stop patterns that should be replaced with AskUserQuestion
rg -n -C3 --type=md 'stop.*wait|wait.*confirm|wait.*user.*respond|pause.*user' workflows/bugfix/.claude/skills/ | grep -v AskUserQuestionRepository: ambient-code/workflows
Length of output: 1932
The assess skill correctly uses AskUserQuestion; update pr/SKILL.md to align with the same gate requirement.
The assess phase at lines 96-106 properly uses the AskUserQuestion tool for its hard gate, meeting the controller requirement. However, pr/SKILL.md lines 305-306 still use plain text "Stop again and wait for the user to confirm" instead of the AskUserQuestion tool. Both phase gates must use AskUserQuestion to trigger platform notifications consistently.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@workflows/bugfix/.claude/skills/assess/SKILL.md` around lines 96 - 106, The
pr phase uses plain text "Stop again and wait for the user to confirm" instead
of invoking the AskUserQuestion tool; update the pr phase SKILL.md to call
AskUserQuestion with the same three options used in the assess skill ("PR `#N`
appears to address this bug — review it instead of starting fresh", "PR `#N` is
related but doesn't fully cover it — continue with assessment", "Not sure if PR
`#N` is relevant — continue with assessment") and enforce the hard gate so the
flow does not proceed until AskUserQuestion returns the user's choice.
Remove "bug fix" framing from the description — the skill is general- purpose PR creation machinery. Rewrite to emphasize what it handles (auth detection, fork workflows, default branch, fallbacks) and why you'd use it over ad-hoc gh pr create, so the runner matches it more effectively when the agent needs to create a PR outside the normal workflow flow. Made-with: Cursor
- pr/SKILL.md: move --name-only before -- separator in git diff command (was being parsed as a pathspec, silently returning full diff output) - summary/SKILL.md: exclude summary.md from artifact discovery to prevent self-ingestion on reruns - assess/SKILL.md: fix stale "Step 6" reference → "Step 7" after renumbering - speedrun/SKILL.md: clarify completion report is only for early stops; /summary is the terminal output for normal runs - test/SKILL.md: qualify call_args example as Python-specific Made-with: Cursor
CodeRabbit review responseAddressed 5 findings in commit 9d9a0fc. Here's why the remaining comments were not acted on:
Grep pattern too broad (assess/SKILL.md) — The PR skill should use AskUserQuestion for fork/auth pauses (assess comment → pr/SKILL.md) — The "stop and wait" instructions in the PR skill are operational pauses during a single phase (e.g., "wait for user to create a fork"), not phase transitions. Go test file glob robustness (test/SKILL.md) — The Speedrun summary note about AskUserQuestion gates — The Formatter command examples (test/SKILL.md) — The |
When GITHUB_TOKEN expires mid-session, gh auth status fails but the platform's git credential helper can still provide fresh tokens. The refresh_credentials MCP tool refreshes the backend but doesn't update the shell env var, so the skill needs to bridge that gap. Added recovery path to Step 0: check for a credential helper, query it for a fresh token, and persist via gh auth login --with-token. Also added the 401 expired token symptom to the error recovery table. Found during ACP Custom Workflow testing. Made-with: Cursor
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@workflows/bugfix/.claude/skills/pr/SKILL.md`:
- Around line 443-446: The current step runs "git rebase
FORK_REMOTE/DEFAULT_BRANCH" as if rebasing the feature branch but Step 5 creates
that feature branch afterwards; change the flow so you first check out the base
branch (DEFAULT_BRANCH), fetch/update FORK_REMOTE, and run "git rebase
FORK_REMOTE/DEFAULT_BRANCH" on DEFAULT_BRANCH before creating the feature branch
in Step 5, ensuring the rebase operates on DEFAULT_BRANCH (not the
yet-to-be-created feature branch).
- Around line 398-401: The workflow diff currently compares
FORK_REMOTE/DEFAULT_BRANCH to the local DEFAULT_BRANCH which may be stale;
update the logic that sets WORKFLOW_DIFF to first fetch the remotes and then
diff against the upstream-tracking ref (use UPSTREAM_REMOTE/DEFAULT_BRANCH)
instead of the local branch, e.g. run git fetch for the relevant remotes and
replace the diff target DEFAULT_BRANCH with UPSTREAM_REMOTE/DEFAULT_BRANCH so
WORKFLOW_DIFF uses FORK_REMOTE/DEFAULT_BRANCH..UPSTREAM_REMOTE/DEFAULT_BRANCH
against .github/workflows/.
In `@workflows/bugfix/.claude/skills/summary/SKILL.md`:
- Line 128: Update the user-facing text in SKILL.md where the sentence reads "If
your summary is more than ~40 lines of markdown, cut it down." to use proper
noun capitalization ("Markdown"); search for that exact phrase and replace
"markdown" with "Markdown" (and fix any other occurrences of the lowercase
"markdown" in the same file to maintain consistent, user-facing capitalization).
In `@workflows/bugfix/.claude/skills/test/SKILL.md`:
- Around line 101-108: Replace the blanket instruction "Run the entire test
suite — unit tests, integration tests, and E2E tests" with a clarifying sentence
that only the commands relevant to this repository's language/tooling should be
run, and update the three example lines (pytest tests/, npm test, go test ./...)
to be labeled as examples for "Python projects", "Node.js projects", and "Go
projects" respectively so CI reviewers don't try to run all toolchains when they
aren't present.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 5a92a2bd-4342-49cd-bfd2-951400d536a8
📒 Files selected for processing (5)
workflows/bugfix/.claude/skills/assess/SKILL.mdworkflows/bugfix/.claude/skills/pr/SKILL.mdworkflows/bugfix/.claude/skills/speedrun/SKILL.mdworkflows/bugfix/.claude/skills/summary/SKILL.mdworkflows/bugfix/.claude/skills/test/SKILL.md
The full suite step listed pytest, npm test, and go test without qualification, which could be read as "run all three." Clarify that only the command matching the project's stack should run. Made-with: Cursor
…te clarity - pr/SKILL.md: diff fork against upstream remote tracking ref instead of potentially stale local branch; add UPSTREAM_REMOTE placeholder - pr/SKILL.md: fix rebase ordering — checkout DEFAULT_BRANCH before rebase so the update happens before the feature branch is created - speedrun/SKILL.md: explicitly note AskUserQuestion gates still apply in speedrun mode - summary/SKILL.md: capitalize "Markdown" as proper noun Made-with: Cursor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@workflows/bugfix/.claude/skills/pr/SKILL.md`:
- Around line 195-203: The unauthenticated git remote detection currently
hardcodes "origin" in the commands extracting repo name and default branch;
change those commands to use the UPSTREAM_REMOTE variable determined in Step 2b
(e.g., replace occurrences of "origin" with "$UPSTREAM_REMOTE" or
"${UPSTREAM_REMOTE}") so the repo and HEAD branch are read from the true
upstream remote instead of the fork; ensure the script references the same
UPSTREAM_REMOTE variable name used earlier and quotes it appropriately in the
git remote get-url and git remote show calls.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 141387d9-f018-44ae-b80a-70afd4dafe42
📒 Files selected for processing (3)
workflows/bugfix/.claude/skills/pr/SKILL.mdworkflows/bugfix/.claude/skills/speedrun/SKILL.mdworkflows/bugfix/.claude/skills/summary/SKILL.md
The git-based repo name and default branch detection hardcoded "origin", which breaks when origin points to the fork. Use the UPSTREAM_REMOTE identified in Step 2b. Made-with: Cursor
There was a problem hiding this comment.
Actionable comments posted: 3
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
workflows/bugfix/.claude/skills/pr/SKILL.md (1)
168-183: 🛠️ Refactor suggestion | 🟠 MajorAdd explicit logic for determining UPSTREAM_REMOTE.
Step 2b lists common patterns but doesn't provide explicit instructions for setting the
UPSTREAM_REMOTEvariable. This leaves the detection logic implementation-dependent.Consider adding explicit detection commands, for example:
# Check which remote points to the upstream org (not the fork) # Assuming UPSTREAM_OWNER was determined in Step 2c UPSTREAM_REMOTE=$(git remote -v | grep -E "github\.com[:/]${UPSTREAM_OWNER}/" | head -1 | awk '{print $1}') # Common fallback: if origin points to upstream org, use it if [ -z "$UPSTREAM_REMOTE" ]; then UPSTREAM_REMOTE="origin" fiThis ensures consistent detection across different AI agent implementations.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@workflows/bugfix/.claude/skills/pr/SKILL.md` around lines 168 - 183, Step 2b describes common remote name patterns but doesn't set UPSTREAM_REMOTE; add explicit detection logic that finds the remote whose URL contains the upstream owner (use the already-determined UPSTREAM_OWNER) and assigns its name to UPSTREAM_REMOTE, falling back to a sensible default such as "origin" if none found; implement this where remotes are inventoried (referencing UPSTREAM_REMOTE and UPSTREAM_OWNER) so downstream steps reliably use the correct upstream remote.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@workflows/bugfix/.claude/skills/pr/SKILL.md`:
- Around line 399-404: The script mixes bare template placeholders (FORK_REMOTE,
UPSTREAM_REMOTE, DEFAULT_BRANCH) with shell variables (e.g., "$FRESH_TOKEN",
WORKFLOW_DIFF) which is inconsistent and will break execution; pick the
shell-variable approach: after detection/or assignment set shell variables
UPSTREAM_REMOTE, FORK_REMOTE and DEFAULT_BRANCH, then update all occurrences to
use the shell-variable form (e.g., git fetch "$FORK_REMOTE", git fetch
"$UPSTREAM_REMOTE" and use "$FORK_REMOTE/$DEFAULT_BRANCH" in the WORKFLOW_DIFF
assignment) so every reference (FORK_REMOTE, UPSTREAM_REMOTE, DEFAULT_BRANCH,
WORKFLOW_DIFF, FRESH_TOKEN) is consistently used as a shell variable.
- Around line 100-128: The recovery flow reads a potential token into
FRESH_TOKEN via git credential fill but then uses it unconditionally; add guards
that check FRESH_TOKEN is non-empty before exporting it, calling gh auth status,
or running gh auth login --with-token so these commands are only invoked when a
token was actually returned. Locate the token extraction (FRESH_TOKEN=$(...git
credential fill...)) and wrap the subsequent export GITHUB_TOKEN, gh auth
status, and gh auth login --with-token invocations in a conditional test that
ensures [ -n "$FRESH_TOKEN" ] (or equivalent) before performing those actions.
- Around line 195-203: The two shell commands that reference the upstream
remote—git remote get-url and git remote show—are using the literal token
UPSTREAM_REMOTE instead of the shell variable; update both invocations to use
the shell variable (e.g., wrap the name in quotes as "$UPSTREAM_REMOTE") so the
variable's value is used safely (preserve the sed/grep/awk parts unchanged),
ensuring commands target the resolved upstream remote.
---
Outside diff comments:
In `@workflows/bugfix/.claude/skills/pr/SKILL.md`:
- Around line 168-183: Step 2b describes common remote name patterns but doesn't
set UPSTREAM_REMOTE; add explicit detection logic that finds the remote whose
URL contains the upstream owner (use the already-determined UPSTREAM_OWNER) and
assigns its name to UPSTREAM_REMOTE, falling back to a sensible default such as
"origin" if none found; implement this where remotes are inventoried
(referencing UPSTREAM_REMOTE and UPSTREAM_OWNER) so downstream steps reliably
use the correct upstream remote.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: b22788ca-3bc3-4bc1-9d85-a9a8c5cbcf12
📒 Files selected for processing (1)
workflows/bugfix/.claude/skills/pr/SKILL.md
| ```bash | ||
| git config --global credential.helper 2>/dev/null | ||
| ``` | ||
|
|
||
| If a credential helper exists (e.g., `/tmp/git-credential-ambient`), query it | ||
| for a fresh token: | ||
|
|
||
| ```bash | ||
| FRESH_TOKEN=$(printf 'protocol=https\nhost=github.com\n\n' | git credential fill 2>/dev/null | grep '^password=' | cut -d= -f2) | ||
| ``` | ||
|
|
||
| If that returns a token, export it and re-check auth: | ||
|
|
||
| ```bash | ||
| export GITHUB_TOKEN="$FRESH_TOKEN" | ||
| gh auth status | ||
| ``` | ||
|
|
||
| **Important:** Each shell invocation gets a fresh environment, so you must | ||
| prepend the export to every subsequent `gh` command, or write the token to | ||
| `~/.config/gh/hosts.yml` so `gh` picks it up natively: | ||
|
|
||
| ```bash | ||
| gh auth login --with-token <<< "$FRESH_TOKEN" | ||
| ``` | ||
|
|
||
| The `gh auth login` approach is preferred — it persists for all subsequent | ||
| `gh` commands without per-command exports. After recovery, re-run the identity | ||
| checks above to set `AUTH_TYPE` and `GH_USER`. |
There was a problem hiding this comment.
🛠️ Refactor suggestion | 🟠 Major
Add conditional check before using FRESH_TOKEN.
The token recovery flow queries the credential helper but doesn't verify the token is non-empty before using it. If the credential helper returns nothing, the subsequent commands will fail with an empty token.
♻️ Proposed fix
```bash
FRESH_TOKEN=$(printf 'protocol=https\nhost=github.com\n\n' | git credential fill 2>/dev/null | grep '^password=' | cut -d= -f2)-If that returns a token, export it and re-check auth:
+If that returns a non-empty token, export it and re-check auth:
+if [ -n "$FRESH_TOKEN" ]; then
+ export GITHUB_TOKEN="$FRESH_TOKEN"
+ gh auth status
+fi
-export GITHUB_TOKEN="$FRESH_TOKEN"
-gh auth statusImportant: Each shell invocation gets a fresh environment, so you must
prepend the export to every subsequent gh command, or write the token to
~/.config/gh/hosts.yml so gh picks it up natively:
+if [ -n "$FRESH_TOKEN" ]; then
+ gh auth login --with-token <<< "$FRESH_TOKEN"
+fi
-gh auth login --with-token <<< "$FRESH_TOKEN"</details>
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against the current code and only fix it if needed.
In @workflows/bugfix/.claude/skills/pr/SKILL.md around lines 100 - 128, The
recovery flow reads a potential token into FRESH_TOKEN via git credential fill
but then uses it unconditionally; add guards that check FRESH_TOKEN is non-empty
before exporting it, calling gh auth status, or running gh auth login
--with-token so these commands are only invoked when a token was actually
returned. Locate the token extraction (FRESH_TOKEN=$(...git credential fill...))
and wrap the subsequent export GITHUB_TOKEN, gh auth status, and gh auth login
--with-token invocations in a conditional test that ensures [ -n "$FRESH_TOKEN"
] (or equivalent) before performing those actions.
</details>
<!-- fingerprinting:phantom:triton:puma -->
<!-- This is an auto-generated comment by CodeRabbit -->
| If `gh` is NOT authenticated, extract from the upstream remote (identified in | ||
| Step 2b as `UPSTREAM_REMOTE`): | ||
|
|
||
| ```bash | ||
| git remote get-url origin | sed -E 's#.*/([^/]+/[^/]+?)(\.git)?$#\1#' | ||
| # Repo name (use UPSTREAM_REMOTE, not necessarily origin — origin may be the fork) | ||
| git remote get-url UPSTREAM_REMOTE | sed -E 's#.*/([^/]+/[^/]+?)(\.git)?$#\1#' | ||
|
|
||
| # Default branch | ||
| git remote show UPSTREAM_REMOTE 2>/dev/null | grep 'HEAD branch' | awk '{print $NF}' |
There was a problem hiding this comment.
Critical: Use proper shell variable syntax for UPSTREAM_REMOTE.
The commands use bare UPSTREAM_REMOTE without shell variable syntax, which will fail when executed. Bash will look for a remote literally named "UPSTREAM_REMOTE" rather than using the variable's value.
This is inconsistent with the file's own usage of proper shell syntax elsewhere (e.g., line 108 uses "$FRESH_TOKEN", line 404 assigns to WORKFLOW_DIFF).
🐛 Proposed fix
```bash
-# Repo name (use UPSTREAM_REMOTE, not necessarily origin — origin may be the fork)
-git remote get-url UPSTREAM_REMOTE | sed -E 's#.*/([^/]+/[^/]+?)(\.git)?$#\1#'
+# Repo name (from upstream remote found in Step 2b)
+git remote get-url "$UPSTREAM_REMOTE" | sed -E 's#.*/([^/]+/[^/]+?)(\.git)?$#\1#'
# Default branch
-git remote show UPSTREAM_REMOTE 2>/dev/null | grep 'HEAD branch' | awk '{print $NF}'
+git remote show "$UPSTREAM_REMOTE" 2>/dev/null | grep 'HEAD branch' | awk '{print $NF}'</details>
**Note:** This was marked as addressed in commit 44f926e, but the shell variable syntax was not applied correctly.
<details>
<summary>🤖 Prompt for AI Agents</summary>
Verify each finding against the current code and only fix it if needed.
In @workflows/bugfix/.claude/skills/pr/SKILL.md around lines 195 - 203, The two
shell commands that reference the upstream remote—git remote get-url and git
remote show—are using the literal token UPSTREAM_REMOTE instead of the shell
variable; update both invocations to use the shell variable (e.g., wrap the name
in quotes as "$UPSTREAM_REMOTE") so the variable's value is used safely
(preserve the sed/grep/awk parts unchanged), ensuring commands target the
resolved upstream remote.
</details>
<!-- fingerprinting:phantom:triton:puma -->
<!-- This is an auto-generated comment by CodeRabbit -->
| git fetch FORK_REMOTE | ||
| git fetch UPSTREAM_REMOTE | ||
|
|
||
| # Check for workflow file differences between fork/main and local main | ||
| # (local main should be synced with upstream) | ||
| WORKFLOW_DIFF=$(git diff fork/main..main -- .github/workflows/ --name-only 2>/dev/null) | ||
| # Compare fork's DEFAULT_BRANCH against upstream's DEFAULT_BRANCH | ||
| # (don't rely on the local branch — it may be stale) | ||
| WORKFLOW_DIFF=$(git diff --name-only FORK_REMOTE/DEFAULT_BRANCH..UPSTREAM_REMOTE/DEFAULT_BRANCH -- .github/workflows/ 2>/dev/null) |
There was a problem hiding this comment.
Systematic inconsistency: Mix of template placeholders and shell variables.
The file mixes two incompatible patterns:
- Template placeholders (bare UPPERCASE):
git fetch FORK_REMOTE,git checkout DEFAULT_BRANCH - Shell variables (with $ syntax):
"$FRESH_TOKEN"(line 108),WORKFLOW_DIFF=$(...)(line 404)
This creates confusion about which commands are executable vs. require manual substitution. For example:
- Line 108 correctly uses
"$FRESH_TOKEN" - Line 404 correctly assigns
WORKFLOW_DIFF=$(...) - But lines 399-400 use bare
FORK_REMOTEandUPSTREAM_REMOTEwhich won't work in bash
The inconsistency makes it unclear whether this file expects:
- AI agents to substitute placeholders before execution, OR
- Users to run commands directly with pre-set shell variables
💡 Recommended approach
Option 1 (Recommended): Use shell variables consistently throughout. Set them in Step 2e after detection:
# After determining values in Step 2b-2c, set shell variables:
UPSTREAM_REMOTE="origin" # or whatever was detected
FORK_REMOTE="fork" # or whatever was detected
DEFAULT_BRANCH="main" # or whatever was detectedThen use "$VAR" syntax in all subsequent commands:
git fetch "$FORK_REMOTE"
git fetch "$UPSTREAM_REMOTE"
WORKFLOW_DIFF=$(git diff --name-only "$FORK_REMOTE/$DEFAULT_BRANCH".."$UPSTREAM_REMOTE/$DEFAULT_BRANCH" -- .github/workflows/ 2>/dev/null)Option 2: Add explicit guidance that ALL_CAPS placeholders must be manually substituted by the AI agent before running commands, and keep the current template style—but then remove the conflicting shell variable usage of FRESH_TOKEN and WORKFLOW_DIFF to maintain consistency.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@workflows/bugfix/.claude/skills/pr/SKILL.md` around lines 399 - 404, The
script mixes bare template placeholders (FORK_REMOTE, UPSTREAM_REMOTE,
DEFAULT_BRANCH) with shell variables (e.g., "$FRESH_TOKEN", WORKFLOW_DIFF) which
is inconsistent and will break execution; pick the shell-variable approach:
after detection/or assignment set shell variables UPSTREAM_REMOTE, FORK_REMOTE
and DEFAULT_BRANCH, then update all occurrences to use the shell-variable form
(e.g., git fetch "$FORK_REMOTE", git fetch "$UPSTREAM_REMOTE" and use
"$FORK_REMOTE/$DEFAULT_BRANCH" in the WORKFLOW_DIFF assignment) so every
reference (FORK_REMOTE, UPSTREAM_REMOTE, DEFAULT_BRANCH, WORKFLOW_DIFF,
FRESH_TOKEN) is consistently used as a shell variable.
1 participant
Summary
Comprehensive update to the bugfix workflow addressing 9 open issues across 6 skill files.
Changes by file
controller/SKILL.md (#81)
AskUserQuestiontool for phase gates instead of prose "stop and wait" — creates a hard structural gate with platform notificationsassess/SKILL.md (#82)
reproduce/SKILL.md (#57)
fix/SKILL.md (#58)
test/SKILL.md (#55, #56, #59)
pr/SKILL.md (#80, #85)
DEFAULT_BRANCHplaceholder: detect from upstream repo instead of hardcodingmain(Bugfix workflow: PR skill assumes main as default branch #85)FORK_REMOTEplaceholder: set based on which remote points to the fork instead of hardcodingfork(Bugfix workflow PR skill: hardcoded fork remote name contradicts Step 4 guidance #80)Test plan
mainas default branch — verify phase gates work with AskUserQuestiondevas default branch — verify PR targets the correct branchCustom Workflow Testing
https://github.com/jwm4/workflows.gitbugfix/comprehensive-workflow-improvementsworkflows/bugfixAddresses #55, #56, #57, #58, #59, #80, #81, #82, #85
Made with Cursor