xonecas.io
64875
Mc

mini-coder / commit 1d30d9a

chore: cleanup ai slop from `terminal-bench`

xonecas2026-04-13 1d30d9a745abd6efb78955e5990c1a375d306d2e parent 07f33ed

4 files changed

terminal-bench/jobs/leaderboard-2026-04-10__16-06-57.launch.log+0−21
@@ -1,21 +0,0 @@
1- 89/89 Mean: 0.674 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 20:25:28 0:00:00Results written to terminal-bench/jobs/leaderboard-2026-04-10__16-06-57/result.json
2- mini-coder (gpt-5.4) on terminal-bench
3-┏━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━┓
4-┃ Metric ┃ Value ┃
5-┡━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━┩
6-│ Agent │ mini-coder (gpt-5.4) │
7-│ Dataset │ terminal-bench │
8-│ Trials │ 87 │
9-│ Errors │ 15 │
10-│ │ │
11-│ Mean │ 0.674 │
12-│ │ │
13-│ Reward Distribution │ │
14-│ reward = 0.0 │ 27 │
15-│ reward = 1.0 │ 60 │
16-│ │ │
17-│ Exception Distribution │ │
18-│ AgentTimeoutError │ 14 │
19-│ VerifierTimeoutError │ 1 │
20-└────────────────────────┴──────────────────────┘
21-
terminal-bench/leaderboard-2026-04-10__16-06-57-failures-rerun-2.prompt-change-analysis.md+0−266
@@ -1,266 +0,0 @@
1-# Leaderboard failures rerun analysis
2-
3-This report compares:
4-
5-- original leaderboard run: `terminal-bench/jobs/leaderboard-2026-04-10__16-06-57`
6-- failure rerun: `terminal-bench/jobs/leaderboard-2026-04-10__16-06-57-failures-rerun-2`
7-
8-Data sources used:
9-
10-- each trial `result.json`
11-- verifier `ctrf.json` and `test-stdout.txt`
12-- agent logs in `agent/mini-coder.ndjson`
13-
14-## Executive summary
15-
16-This rerun covered **29 tasks that had failed in the original leaderboard run**.
17-
18-Outcome on the rerun set:
19-
20-- **12 now pass**
21-- **16 still fail**
22-- **1 timed out before verifier completion** (`video-processing`)
23-
24-The fail -> pass tasks are:
25-
26-- `adaptive-rejection-sampler`
27-- `dna-insert`
28-- `extract-elf`
29-- `largest-eigenval`
30-- `llm-inference-batching-scheduler`
31-- `mailman`
32-- `overfull-hbox`
33-- `polyglot-c-py`
34-- `polyglot-rust-c`
35-- `qemu-alpine-ssh`
36-- `torch-tensor-parallelism`
37-- `tune-mjcf`
38-
39-The improvement is real. The strongest signal is that **two tasks now pass even though the agent still timed out**:
40-
41-- `mailman`
42-- `tune-mjcf`
43-
44-That means the required artifact was already good enough before the run ended. This matches the intended prompt change: produce the required output earlier instead of saving it for late in the run.
45-
46-## What improved
47-
48-## 1. More tasks now cross the verifier line
49-
50-Several tasks moved from obvious near-miss territory to clean pass territory:
51-
52-| task | original | rerun |
53-| ---------------------------------- | ------------------ | ------------------------------- |
54-| `llm-inference-batching-scheduler` | 1/6 tests, timeout | 6/6 tests, pass |
55-| `polyglot-rust-c` | 0/1 tests, timeout | 1/1 tests, pass |
56-| `qemu-alpine-ssh` | 0/1 tests, timeout | 1/1 tests, pass |
57-| `mailman` | 1/3 tests, timeout | 3/3 tests, pass despite timeout |
58-| `tune-mjcf` | 3/4 tests, timeout | 4/4 tests, pass despite timeout |
59-| `adaptive-rejection-sampler` | 8/9 tests | 9/9 tests |
60-| `extract-elf` | 1/2 tests | 2/2 tests |
61-| `largest-eigenval` | 2/3 tests | 3/3 tests |
62-| `torch-tensor-parallelism` | 2/3 tests | 3/3 tests |
63-| `overfull-hbox` | 3/4 tests | 4/4 tests |
64-
65-This is not just a small shift in average behavior. Multiple tasks crossed from partial verifier progress to full pass.
66-
67-## 2. The agent appears to verify a bit more, and earlier
68-
69-Using a rough heuristic over shell commands, the rerun subset shows more verifier/test-like shell usage:
70-
71-- original rerun-subset baseline: **8** test/verifier-like shell calls
72-- rerun: **18** test/verifier-like shell calls
73-
74-Within the rerun itself, the passing tasks also look more verifier-oriented than the still-failing tasks:
75-
76-- tasks that improved to pass: **0.92** test-like shell calls on average
77-- tasks that still failed: **0.41** test-like shell calls on average
78-
79-This is only a heuristic, but it lines up with the visible task outcomes: more of the successful runs found the verifier boundary and iterated against it.
80-
81-## 3. Some runs now create a passing artifact before the timeout wall
82-
83-The clearest evidence is again `mailman` and `tune-mjcf`, which both still ended with `AgentTimeoutError` but now score `1.0`.
84-
85-That is exactly the failure mode the prompt change was meant to attack: runs that do enough good work to pass, but only if the required artifact lands before the timeout.
86-
87-## Tool awkwardness: what changed
88-
89-## `edit`
90-
91-`edit` no longer looks like a major source of friction.
92-
93-Across the 29-task rerun subset:
94-
95-- **79** edit calls total
96-- only **2** edit errors total
97-
98-Those two edit errors were straightforward exact-match failures:
99-
100-- `caffe-cifar-10` — old text matched multiple locations
101-- `install-windows-3.11` — old text not found
102-
103-Important contrast with the original run: the earlier internal `edit` crash seen in `gpt2-codegolf` (`undefined is not an object`) did **not** reappear here.
104-
105-Conclusion: from the agent's perspective, `edit` is mostly behaving cleanly now.
106-
107-## `shell`
108-
109-Raw non-zero shell exits are almost unchanged:
110-
111-- original rerun-subset baseline: **68** non-zero shell results
112-- rerun: **67** non-zero shell results
113-
114-So the improvement is **not** “the agent suddenly stopped making shell mistakes”. The more useful change is in the type of shell awkwardness.
115-
116-### The earlier `printf` compatibility issue appears fixed
117-
118-The original run contained several `printf: Illegal option --` failures in agent-authored shell commands.
119-
120-I found those in the original logs, including `mailman`, `query-optimize`, and other tasks. I found **no occurrences** of that issue in the rerun logs.
121-
122-That is a direct positive signal for the shell compatibility normalization work.
123-
124-### The remaining shell awkwardness is mostly environment assumptions
125-
126-The recurring shell failures in the rerun are dominated by assumptions about available binaries rather than mini-coder shell semantics.
127-
128-Recurring examples in the rerun subset:
129-
130-- `python` not found — **8 tasks**
131-- `file` not found — **4 tasks**
132-- `No module named pip` — **3 tasks**
133-- `/usr/bin/time` not found — **2 tasks**
134-- `g++` not found — **2 tasks**
135-
136-Tasks hit by `python` not found:
137-
138-- `dna-assembly`
139-- `dna-insert`
140-- `extract-moves-from-video`
141-- `mcmc-sampling-stan`
142-- `overfull-hbox`
143-- `qemu-alpine-ssh`
144-- `torch-pipeline-parallelism`
145-- `torch-tensor-parallelism`
146-
147-This matters because some of these tasks still passed anyway. So these failures are real friction, but they are not the same class as the earlier shell-tool conflict. The agent is still making environment assumptions, but mini-coder itself is less obviously fighting the agent.
148-
149-## Remaining failures: how they still fail
150-
151-## A. Timed out without producing the required artifact
152-
153-These are still the clearest contract-first misses.
154-
155-| task | rerun verifier | main failure signal |
156-| -------------------------- | -------------------- | ------------------------------------ |
157-| `crack-7z-hash` | 0/2 | `/app/solution.txt` missing |
158-| `db-wal-recovery` | 0/7 | `/app/recovered.json` missing |
159-| `extract-moves-from-video` | 0/2 | `/app/solution.txt` missing |
160-| `make-doom-for-mips` | 0/3 | `/tmp/frame.bmp` missing |
161-| `video-processing` | verifier not reached | timed out before verifier completion |
162-
163-This is still the biggest remaining bucket of lost score.
164-
165-## B. Timed out with partial progress, but not enough to pass
166-
167-These runs clearly did real work, but did not land on the exact contract in time.
168-
169-| task | rerun verifier | main failure signal |
170-| ---------------------- | -------------- | ------------------------------------------------------------ |
171-| `caffe-cifar-10` | 3/6 | `caffe.bin` missing; solver not set to CPU mode |
172-| `install-windows-3.11` | 2/4 | no VGA config; keyboard/F1 monitor interaction still failing |
173-| `train-fasttext` | 0/2 | accuracy too low and model too large |
174-
175-`caffe-cifar-10` is especially worth noting: it improved from “timeout before verifier signal” in the original run to **3/6 passing tests** in the rerun, so it is clearly closer, but still not there.
176-
177-## C. Near misses: one bad detail blocks an otherwise strong run
178-
179-These are the rerun failures that already look close to passing.
180-
181-| task | rerun verifier | main failure signal |
182-| ---------------------------- | -------------- | ------------------------------------------------------------- |
183-| `headless-terminal` | 6/7 | interactive path breaks inside `pyte` on private SGR handling |
184-| `mcmc-sampling-stan` | 5/6 | expected sampling log/messages missing |
185-| `make-mips-interpreter` | 2/3 | output image similarity 0.8456, needs 0.95 |
186-| `torch-pipeline-parallelism` | 2/3 | backward mismatch in one layer/microbatch |
187-
188-These are not broad failures. They are verifier-focused misses on one remaining edge or correctness detail.
189-
190-## D. Wrong solution shape or algorithm mismatch
191-
192-These failures do not look tool-related. They look like the agent produced the wrong implementation or the wrong tradeoff.
193-
194-| task | rerun verifier | main failure signal |
195-| --------------------- | -------------- | -------------------------------------------------------- |
196-| `dna-assembly` | 0/1 | primer constraints wrong |
197-| `filter-js-from-html` | 0/2 | still misses many XSS cases and also modifies clean HTML |
198-| `gpt2-codegolf` | 0/1 | wrong output |
199-| `raman-fitting` | 1/3 | fitted peaks are materially wrong |
200-| `sam-cell-seg` | 1/9 | run script fails; expected output CSV missing |
201-
202-`filter-js-from-html` is the clearest example here. The rerun log shows a lot of detailed work and many tool calls, but the verifier still reports both:
203-
204-- the sanitizer misses large classes of attack vectors
205-- the sanitizer changes clean HTML that should remain unchanged
206-
207-That is a solution-fit issue, not a tool-usage issue.
208-
209-## Regressions and unstable cases
210-
211-Not everything improved.
212-
213-A few tasks got worse, or at least less stable:
214-
215-| task | original | rerun |
216-| ---------------------- | --------- | ----------------------------- |
217-| `db-wal-recovery` | 5/7 tests | 0/7 tests, timeout |
218-| `install-windows-3.11` | 3/4 tests | 2/4 tests |
219-| `video-processing` | 4/5 tests | verifier not reached, timeout |
220-
221-A few others stayed stuck at roughly the same near-miss level:
222-
223-- `headless-terminal` — 6/7 -> 6/7
224-- `make-mips-interpreter` — 2/3 -> 2/3
225-- `torch-pipeline-parallelism` — 2/3 -> 2/3
226-- `raman-fitting` — 1/3 -> 1/3
227-- `sam-cell-seg` — 1/9 -> 1/9
228-
229-So the gains are real, but they are not universal.
230-
231-## Overall conclusion
232-
233-The rerun supports the main hypothesis behind the recent prompt and tool work:
234-
235-- the agent is more likely to produce the required output early enough to matter
236-- the agent is a bit more likely to discover verifier boundaries and iterate against them
237-- the earlier `printf` shell awkwardness looks gone
238-- `edit` now looks low-friction most of the time
239-
240-The remaining misses are now less about mini-coder fighting the agent, and more about three buckets:
241-
242-1. still timing out before writing the exact required artifact
243-2. near-miss implementations with one remaining verifier-visible defect
244-3. environment assumptions in shell usage (`python`, `file`, `pip`, `g++`, `/usr/bin/time`)
245-
246-That is a healthier failure profile than the original run.
247-
248-## Ranked narrow changes to discuss next
249-
250-These are ranked discussion candidates, based on the rerun data above.
251-
252-1. **Add a narrow shell compatibility rewrite from `python` -> `python3` when `python` is absent and the rewrite is unambiguous.**
253- - Why this ranks first: `python`-missing shell failures appeared in **8 rerun tasks**, including some otherwise strong runs. This is a narrow, repeated, machine-fixable pattern.
254-
255-2. **Add an explicit prompt rule to probe tool availability before assuming non-core binaries exist.**
256- - Example pattern: `command -v file`, `command -v g++`, `command -v python3`, `python3 -m pip --version`.
257- - Why: the rerun still shows repeated assumptions about `file`, `pip`, `g++`, and `/usr/bin/time`.
258-
259-3. **Strengthen the prompt rule that if the task names a required output path, the agent should create that exact artifact early, then iterate.**
260- - Why: the biggest remaining zero-score failures still end with missing required files like `/app/solution.txt`, `/app/recovered.json`, or `/tmp/frame.bmp`.
261-
262-4. **Strengthen the prompt rule to run the narrowest verifier or test as soon as there is a plausible first artifact, instead of delaying the first verifier run.**
263- - Why: the improved tasks correlate with more verifier-like shell usage, and many remaining failures still look under-verified until late.
264-
265-5. **Add a prompt rule for near-miss mode: once most tests are passing, stop broad exploration and focus only on the failing assertions.**
266- - Why: several rerun failures are already at 5/6, 6/7, or 2/3 and look like they lost time to broader work instead of concentrating on the final failing detail.
terminal-bench/mini_coder_local_agent.py+0−72
@@ -1,72 +0,0 @@
1-from __future__ import annotations
2-
3-import subprocess
4-import tempfile
5-from pathlib import Path
6-
7-from harbor.environments.base import BaseEnvironment
8-
9-from mini_coder_agent import MiniCoderAgent
10-
11-
12-class MiniCoderLocalAgent(MiniCoderAgent):
13- """Harbor installed-agent wrapper for the current local mini-coder checkout."""
14-
15- def __init__(self, *args, version: str = "local-checkout", **kwargs):
16- super().__init__(*args, version=version, **kwargs)
17- self._repo_root = Path(__file__).resolve().parent.parent
18-
19- @staticmethod
20- def name() -> str:
21- return "mini-coder-local"
22-
23- def _pack_local_checkout(self, output_dir: Path) -> Path:
24- tarball_name = "mini-coder-local.tgz"
25- tarball_path = output_dir / tarball_name
26- try:
27- result = subprocess.run(
28- [
29- "bun",
30- "pm",
31- "pack",
32- "--ignore-scripts",
33- "--filename",
34- str(tarball_path),
35- ],
36- cwd=self._repo_root,
37- check=False,
38- capture_output=True,
39- text=True,
40- )
41- except FileNotFoundError as error:
42- raise RuntimeError(
43- "bun is required on the host to pack the current mini-coder checkout"
44- ) from error
45-
46- if result.returncode != 0:
47- raise RuntimeError(
48- "Failed to pack the current mini-coder checkout for Harbor.\n"
49- f"stdout:\n{result.stdout}\n"
50- f"stderr:\n{result.stderr}"
51- )
52- if not tarball_path.is_file():
53- raise RuntimeError(
54- f"bun pm pack did not create the expected tarball: {tarball_path}"
55- )
56- return tarball_path
57-
58- async def install(self, environment: BaseEnvironment) -> None:
59- await self._install_system_dependencies(environment)
60- await self._install_bun(environment)
61-
62- with tempfile.TemporaryDirectory(
63- prefix="mini-coder-terminal-bench-"
64- ) as output_dir:
65- tarball_path = self._pack_local_checkout(Path(output_dir))
66- container_tarball_path = f"/tmp/{tarball_path.name}"
67- await environment.upload_file(tarball_path, container_tarball_path)
68- await self._install_global_package(
69- environment,
70- container_tarball_path,
71- ignore_scripts=True,
72- )
terminal-bench/readme.md+0−218
@@ -1,218 +0,0 @@
1-# terminal-bench benchmark workspace
2-
3-This directory contains Harbor wrappers and job output for benchmarking `mini-coder` on Terminal-Bench 2.0.
4-
5-Everything here stays under `terminal-bench/` so benchmark work stays separate from production mini-coder code.
6-
7-## Goal
8-
9-Keep two workflows side by side:
10-
11-- `mini_coder_agent.py` — leaderboard-style runs against published `mini-coder@latest`
12-- `mini_coder_local_agent.py` — pre-commit runs against the current local packaged checkout
13-
14-Both wrappers use the current local mini-coder config:
15-
16-- model: `openai-codex/gpt-5.4`
17-- auth source: `~/.config/mini-coder/auth.json`
18-- settings source: `~/.config/mini-coder/settings.json`
19-- effort source: `~/.config/mini-coder/settings.json`
20-
21-At the time of writing, `npm view mini-coder version dist-tags --json` reports `latest = 0.5.2`.
22-
23-## Files
24-
25-- `mini_coder_agent.py` — Harbor installed-agent wrapper for published `mini-coder@latest`
26-- `mini_coder_local_agent.py` — Harbor installed-agent wrapper for the current local packaged checkout
27-- `jobs/` — Harbor job outputs
28-- `readme.md` — this runbook
29-
30-## What each wrapper does
31-
32-### `mini_coder_agent.py`
33-
34-Per trial it:
35-
36-1. installs system deps: `curl`, `unzip`, `ca-certificates`, `bash`, `git`, `python3`, `ripgrep`
37-2. installs Bun
38-3. installs published `mini-coder@latest`
39-4. finds the container user's home directory
40-5. uploads copies of:
41- - `~/.config/mini-coder/settings.json`
42- - `~/.config/mini-coder/auth.json`
43-6. runs:
44- - `mc -p <task instruction>`
45-7. writes logs to:
46- - `/logs/agent/mini-coder.ndjson`
47- - `/logs/agent/mini-coder.stderr.txt`
48-
49-### `mini_coder_local_agent.py`
50-
51-Per trial it:
52-
53-1. installs the same system deps and Bun
54-2. runs `bun pm pack --ignore-scripts` on the current repo checkout on the host
55-3. uploads that tarball into the Harbor container
56-4. installs `mini-coder` from the uploaded tarball
57-5. uploads host copies of `settings.json` and `auth.json`
58-6. runs:
59- - `mc -p <task instruction>`
60-7. writes logs to:
61- - `/logs/agent/mini-coder.ndjson`
62- - `/logs/agent/mini-coder.stderr.txt`
63-
64-Use the local wrapper when you want Terminal-Bench signal on the current diff before committing. Use the published wrapper for leaderboard-comparable runs.
65-
66-## Requirements
67-
68-Before running Harbor:
69-
70-- `harbor` must be installed and on `PATH`
71-- local mini-coder auth must be valid
72-- local mini-coder settings must exist
73-- `bun` must be installed on the host for local-checkout runs
74-
75-Quick checks:
76-
77-```bash
78-command -v harbor
79-command -v bun
80-ls -l ~/.config/mini-coder/settings.json ~/.config/mini-coder/auth.json
81-npm view mini-coder version dist-tags --json
82-```
83-
84-## Verify the wrappers import
85-
86-```bash
87-PYTHONPATH=$PWD/terminal-bench \
88- ~/.local/share/pipx/venvs/harbor/bin/python -m py_compile \
89- terminal-bench/mini_coder_agent.py \
90- terminal-bench/mini_coder_local_agent.py
91-```
92-
93-## Recommended smoke runs
94-
95-Run these easy tasks first before slower tasks.
96-
97-### Published `@latest`: `fix-git`
98-
99-```bash
100-PYTHONPATH=$PWD/terminal-bench harbor run -y \
101- -d terminal-bench@2.0 \
102- --agent-import-path mini_coder_agent:MiniCoderAgent \
103- -m openai-codex/gpt-5.4 \
104- -i fix-git -l 1 \
105- -k 1 -n 1 \
106- -o terminal-bench/jobs
107-```
108-
109-### Published `@latest`: `prove-plus-comm`
110-
111-```bash
112-PYTHONPATH=$PWD/terminal-bench harbor run -y \
113- -d terminal-bench@2.0 \
114- --agent-import-path mini_coder_agent:MiniCoderAgent \
115- -m openai-codex/gpt-5.4 \
116- -i prove-plus-comm -l 1 \
117- -k 1 -n 1 \
118- -o terminal-bench/jobs
119-```
120-
121-### Current local checkout: `fix-git`
122-
123-```bash
124-PYTHONPATH=$PWD/terminal-bench harbor run -y \
125- -d terminal-bench@2.0 \
126- --agent-import-path mini_coder_local_agent:MiniCoderLocalAgent \
127- -m openai-codex/gpt-5.4 \
128- -i fix-git -l 1 \
129- -k 1 -n 1 \
130- -o terminal-bench/jobs
131-```
132-
133-### Current local checkout: generic one-task command
134-
135-```bash
136-PYTHONPATH=$PWD/terminal-bench harbor run -y \
137- -d terminal-bench@2.0 \
138- --agent-import-path mini_coder_local_agent:MiniCoderLocalAgent \
139- -m openai-codex/gpt-5.4 \
140- -i <task-id> -l 1 \
141- -k 1 -n 1 \
142- -o terminal-bench/jobs
143-```
144-
145-If you want to keep published and local-checkout jobs separate, use different `-o` directories.
146-
147-## Inspecting results
148-
149-### Show the latest Harbor job tree
150-
151-```bash
152-JOB=$(ls -dt terminal-bench/jobs/* | head -n1)
153-find "$JOB" -maxdepth 4 -type f | sort
154-```
155-
156-### Show the latest top-level result summary
157-
158-```bash
159-JOB=$(ls -dt terminal-bench/jobs/* | head -n1)
160-sed -n '1,220p' "$JOB"/result.json
161-```
162-
163-### Show verifier output from the latest trial
164-
165-```bash
166-JOB=$(ls -dt terminal-bench/jobs/* | head -n1)
167-sed -n '1,220p' "$JOB"/*/verifier/test-stdout.txt
168-sed -n '1,40p' "$JOB"/*/verifier/reward.txt
169-```
170-
171-### Show mini-coder logs from the latest trial
172-
173-```bash
174-JOB=$(ls -dt terminal-bench/jobs/* | head -n1)
175-sed -n '1,160p' "$JOB"/*/agent/mini-coder.stderr.txt
176-sed -n '1,160p' "$JOB"/*/agent/mini-coder.ndjson
177-```
178-
179-### Show the latest trial exception, if any
180-
181-```bash
182-JOB=$(ls -dt terminal-bench/jobs/* | head -n1)
183-sed -n '1,220p' "$JOB"/*/exception.txt
184-```
185-
186-## Current benchmark context
187-
188-Latest full leaderboard-style run kept for inspection:
189-
190-- `terminal-bench/jobs/leaderboard-2026-04-10__16-06-57`
191-- score: about **67.4%**
192-- pattern: many failures were near misses; the biggest buckets were exact-contract misses and runs that explored too long before writing the required artifact
193-
194-Working lessons for the next runs:
195-
196-- Read machine-checkable acceptance criteria early: `/tests`, verifier scripts, eval scripts, examples, expected-output files.
197-- Identify the exact contract up front: required file paths, names, signatures, stdout/output shape, and forbidden extra artifacts.
198-- Create the required artifact early, then iterate. Some timed-out runs can still pass if the file is already correct enough.
199-- Prefer the smallest verifier-passing solution over a broader or more impressive implementation.
200-- Before stopping, run a targeted exactness check: required files exist, names/signatures match, output format matches, and no extra artifacts were left behind.
201-
202-Prompt changes prepared after reviewing this run:
203-
204-- base prompt now emphasizes contract-first execution, early verifier/test discovery, early artifact creation, minimal passing solutions, and exact final verification
205-- `shell` tool description now explicitly points the agent toward tests/verifiers/examples and required outputs
206-- `edit` tool description now explicitly frames the tool as writing the exact final file content the task requires
207-
208-Important:
209-
210-- `mini_coder_agent.py` benchmarks the published npm package
211-- `mini_coder_local_agent.py` benchmarks whatever `bun pm pack --ignore-scripts` would install from the current local checkout
212-
213-## Notes
214-
215-- Both wrappers copy host mini-coder auth/settings into the container; they do not mount or mutate the host files.
216-- The task image supplies task-specific toolchains. The wrappers only add generic helpers like `git`, `python3`, and `ripgrep`.
217-- For leaderboard-comparable runs, keep using `mini_coder_agent.py` and do **not** change timeouts or resources.
218-- If `openai-codex` auth expires, ask the user to refresh it locally with mini-coder before rerunning.