executable-issue-resolution

# Skill: executable-issue-resolution ## 1. Capability Definition & Real Case * **Professional Definition**: The ability to resolve a concrete repository issue by reading the issue description, locating the relevant program state, editing code, executing the affected software, and validating that the behavior now satisfies the requested functional outcome without introducing regressions. * **Dimension Hierarchy**: Repository Maintenance and Repair->Issue-Driven Repair->executable-issue-resolution ### Real Case **[Case 1]** * **Initial Environment**: A Python repository contains a data export module, an issue report describing that CSV exports silently drop timezone offsets, and a test suite with one newly failing serialization test plus several passing regression tests. * **Real Question**: Fix the export logic so timestamps retain timezone information in generated CSV files. * **Real Trajectory**: Inspect the serializer used by the export path, trace how datetime objects are converted to strings, modify the formatting branch to preserve aware datetimes, and run the targeted export test plus the broader export test file. * **Real Answer**: The export path now emits ISO-8601 timestamps with offsets for timezone-aware values, and the failing test passes without breaking other export behavior. * **Why this demonstrates the capability**: The task is a canonical executable issue-resolution problem because the issue description alone is insufficient: the agent must inspect the code path, bind the textual symptom to the concrete serialization branch, edit the implementation, and verify the fix against executable evidence. --- **[Case 2]** * **Initial Environment**: A repository for a web service contains an issue reporting that deleted users still appear in cached profile lookups. The codebase includes a cache layer, a delete-user service, and integration tests that fail after deletion. * **Real Question**: Ensure profile cache entries are invalidated when a user account is deleted. * **Real Trajectory**: Search for the delete-account flow, locate cache read and write helpers, add invalidation in the deletion path, and execute the deletion integration test followed by the profile-service regression tests. * **Real Answer**: Deleting an account removes the corresponding cached profile entry, so later reads no longer return stale data. * **Why this demonstrates the capability**: This demonstrates executable issue resolution because the agent must connect an externally described symptom to repository internals, patch the lifecycle logic at the right location, and confirm success by running code rather than by superficial diff similarity. ## Pipeline Execution Instructions To synthesize data for this capability, you must strictly follow a 3-phase pipeline. **Do not hallucinate steps.** Read the corresponding reference file for each phase sequentially: 1. **Phase 1: Environment Exploration** Read the exploration guidelines to discover raw knowledge seeds: `references/EXPLORATION.md` 2. **Phase 2: Trajectory Selection** Once Phase 1 is complete, read the selection criteria to evaluate the trajectory: `references/SELECTION.md` 3. **Phase 3: Data Synthesis** Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data: `references/SYNTHESIS.md`