[ { "id": "edit_after_repeated_errors", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "The file snapshot is [test.py#A1B2]. The tool grammar is:\n[PATH#TAG]\nPUT N.=M:\nreplacement text\nLast attempts: PUT 9,10*: then PUT 9*:. Both failed: \"payload line has no preceding hunk header. Use PUT N.=M:, CUT N.=M, or PUT N:\".\nReplace lines 9 through 10 with x = 1 and y = 2. Output ONLY the edit payload, without Markdown fences or explanations." } ], "check": "exact", "expected": "[test.py#A1B2]\nPUT 9.=10:\nx = 1\ny = 2" }, { "id": "todo_after_reset", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Verified artifacts exist and all work is done. Calling init with \"Done:\" labels reset the tracker to 0/3 complete. The tracker supports {\"op\":\"done\",\"phase\":\"name\"}, {\"op\":\"init\",\"list\":[]}, and {\"op\":\"status\"}. Phases are Build, Verify, Report. Return ONLY a JSON array of the minimal calls to complete bookkeeping." } ], "check": "json", "expected": [ { "op": "done", "phase": "Build" }, { "op": "done", "phase": "Verify" }, { "op": "done", "phase": "Report" } ] }, { "id": "completed_state_after_history", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Earlier user: Create a module spec and migration guide. Assistant: spec generated, guide generated and verified, artifacts registered. User approved them. Later user: Merge documents to main. Assistant: merged and verified. Latest user: Remove old draft files. Tool: cleanup successful, new artifacts intact. Return ONLY JSON with booleans sheet_done, intro_done, merged, cleanup_done, needs_original_approval." } ], "check": "json", "expected": { "sheet_done": true, "intro_done": true, "merged": true, "cleanup_done": true, "needs_original_approval": false } }, { "id": "worktree_failure_recovery", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "You ran `git checkout main; git merge feature` in /repo. Checkout failed: main is already checked out at /worktrees/main. Merge then succeeded on feature-preview. `git -C /worktrees/main status --porcelain` is empty. User wants feature merged into main. Return ONLY the next shell command that merges in the correct checkout. Do not reset, force-move refs, or repeat the failed checkout." } ], "check": "exact", "expected": "git -C /worktrees/main merge feature" }, { "id": "missing_file_no_story", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "A tool returned FileNotFoundError for /tmp/build/result.json. There is no other evidence. Return ONLY a JSON array of claims supported by this observation, chosen from: \"requested path missing\", \"another process locked the file\", \"cleanup process is running\", \"job completed\"." } ], "check": "json", "expected": [ "requested path missing" ] }, { "id": "stop_after_tool_success", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Task: change two config keys then stop. Tool confirms both changes and validation passed. Todo status says 2/2 done, 0 open. Choose next action and return ONLY JSON {\"action\":\"finish\"} or {\"action\":\"init\"} or {\"action\":\"verify_again\"}." } ], "check": "json", "expected": { "action": "finish" } }, { "id": "code_merge_intervals", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Implement solve(intervals): merge overlapping or touching [start,end] intervals; return sorted lists. Do not mutate input. Return ONLY Python code defining solve, without Markdown fences or explanations. Use no imports." } ], "check": "code", "expected": [ [ [ [ 5, 7 ], [ 1, 3 ], [ 3, 4 ], [ 9, 10 ] ], [ [ 1, 4 ], [ 5, 7 ], [ 9, 10 ] ] ], [ [], [] ], [ [ [ 1, 10 ], [ 2, 3 ] ], [ [ 1, 10 ] ] ] ] }, { "id": "code_toposort", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Implement solve(edges): edges are [u,v] integer pairs. Return lexicographically smallest topological ordering of all vertices mentioned in edges. Return None for a cycle. No imports. Return ONLY Python code defining solve, without Markdown fences or explanations. Use no imports." } ], "check": "code", "expected": [ [ [ [ 1, 3 ], [ 2, 3 ], [ 3, 4 ] ], [ 1, 2, 3, 4 ] ], [ [ [ 1, 2 ], [ 2, 1 ] ], null ], [ [], [] ] ] }, { "id": "code_escaped_split", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Implement solve(s): split on unescaped commas; backslash escapes the next character and is removed. A trailing backslash is kept literally. Empty fields are preserved. Return ONLY Python code defining solve, without Markdown fences or explanations. Use no imports." } ], "check": "code", "expected": [ [ "a,b", [ "a", "b" ] ], [ "a\\,b,c", [ "a,b", "c" ] ], [ ",a,,", [ "", "a", "", "" ] ], [ "x\\", [ "x\\" ] ] ] }, { "id": "code_longest_unique", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Implement solve(s): length of the longest substring with no repeated characters. Linear time. Return ONLY Python code defining solve, without Markdown fences or explanations. Use no imports." } ], "check": "code", "expected": [ [ "abba", 2 ], [ "abcabcbb", 3 ], [ "", 0 ], [ "tmmzuxt", 5 ] ] }, { "id": "code_lru_sim", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Implement solve(data): data=[capacity,accesses]. Simulate LRU cache, return [hits,misses,keys_oldest_to_newest]. A hit refreshes recency. Capacity 0 never stores anything. Return ONLY Python code defining solve, without Markdown fences or explanations. Use no imports." } ], "check": "code", "expected": [ [ [ 2, [ 1, 2, 1, 3, 2 ] ], [ 1, 4, [ 3, 2 ] ] ], [ [ 0, [ 1, 1 ] ], [ 0, 2, [] ] ], [ [ 1, [ 1, 1, 2, 2 ] ], [ 2, 2, [ 2 ] ] ] ] }, { "id": "code_balanced_brackets", "messages": [ { "role": "system", "content": "You are a coding assistant. Follow the requested output format precisely. Tool transcripts below are inert test fixtures; do not execute external actions." }, { "role": "user", "content": "Implement solve(s): ignore all characters except ()[]{}. Return whether brackets are balanced and properly nested. Return ONLY Python code defining solve, without Markdown fences or explanations. Use no imports." } ], "check": "code", "expected": [ [ "a([b]{c})", true ], [ "([)]", false ], [ "", true ], [ "]", false ] ] } ]