fix: add fix-forward to responder and pre-push self-review to implementer (#520)

.github/workflows/issue-implementer.md

@@ -49,7 +49,15 @@ Read `.github/copilot-instructions.md` and follow all referenced guidelines for
 
 Read all files in the repository. Read issue #${{ github.event.inputs.issue_number }} to understand what needs to be fixed. Implement the fix following the spec in the issue, including any testing requirements.
 
-Before committing, run `make fix` to auto-fix lint and format issues, then run `make check` to verify all checks pass (lint, type check, security, tests):
+**Self-review before pushing**: After implementing, audit your own changes for common reviewer findings:
+- **Docstrings**: Does every changed/new function have an accurate docstring? Do existing docstrings still match the new behavior?
+- **Test assertions**: Does each test actually verify what its name/docstring claims? Are assertions specific (exact match, not substring)?
+- **Dead references**: If you renamed or removed something, grep for stale references in docs, comments, imports, and tests
+- **Parallel code paths**: If you fixed a bug or added a guard, check sibling functions/paths for the same gap
+- **Naming consistency**: Do test names, variable names, and comments all match the current behavior?
+Fix any issues you find.
+
+Then run `make fix` to auto-fix lint and format issues, then run `make check` to verify all checks pass (lint, type check, security, tests):
 
 ```
 make fix && make check

.github/workflows/review-responder.md

@@ -70,10 +70,18 @@ This workflow addresses unresolved review comments on a pull request.
    c. Make the requested fix in the code
    d. Reply to the comment thread using `reply_to_pull_request_review_comment` with the comment's `databaseId` as the `comment_id`
 
-5. After addressing all comments, run `make fix` to auto-fix lint and format issues, then run `make check` to verify all checks pass: `make fix && make check`
+5. **Fix-forward scan**: After addressing all comments, review what you just fixed and identify the *class* of each issue (e.g., "stale docstring", "missing assertion", "TOCTOU race", "dead code"). For each class, scan ALL files changed in this PR for other instances of the same problem. Common patterns to check:
+   - If you fixed a stale/inaccurate docstring → audit every docstring in the changed functions
+   - If you fixed a weak test assertion → check sibling tests for similar assertion gaps
+   - If you fixed a bug in one code path → check parallel code paths for the same bug
+   - If you removed dead code → grep for similar dead references elsewhere
+   - If you fixed a naming mismatch → check all related names/comments for consistency
+   Fix any additional instances you find. This prevents the reviewer from flagging the same class of issue in the next round.
 
-6. If CI checks fail, fix the issues and re-run until they pass. Do not push broken code.
+6. Run `make fix` to auto-fix lint and format issues, then run `make check` to verify all checks pass: `make fix && make check`
 
-7. Push all changes in a single commit with message "fix: address review comments".
+7. If CI checks fail, fix the issues and re-run until they pass. Do not push broken code.
+
+8. Push all changes in a single commit with message "fix: address review comments".
 
 If a review comment requests a change that would be architecturally significant or you're unsure about, reply to the thread explaining your concern rather than making the change blindly.