Leonxlnx/unlazy

- [#2](https://github.com/Leonxlnx/unlazy/pull/2): re-verification, parser diagnostics, CRLF preservation, evidence insertion, fenced-example handling, and validation ideas

- [#3](https://github.com/Leonxlnx/unlazy/pull/3): the explicit-file positional fix

- [#5](https://github.com/Leonxlnx/unlazy/pull/5): rolling dispatch and bounded `--jobs`

- [#8](https://github.com/Leonxlnx/unlazy/pull/8): stable hook identification and moved-install repair

- [#9](https://github.com/Leonxlnx/unlazy/pull/9): explicit approval for executable checks

- [#10](https://github.com/Leonxlnx/unlazy/pull/10): scoped pipelines, shared parsing, ownership leases, session routing, and the first regression suite

- [#14](https://github.com/Leonxlnx/unlazy/pull/14): the COLM 2026 test-time-scaling source

- [#15](https://github.com/Leonxlnx/unlazy/pull/15): negative controls, supplied-number measurement, and manual-gate review guidance; the single-run risk observation is intentionally not generalized

- [#17](https://github.com/Leonxlnx/unlazy/pull/17): the read-only gate-quality linter, JSON output, strict mode, documentation, and regression harness by Daz Alderson

- [#18](https://github.com/Leonxlnx/unlazy/pull/18): atomic native dispatch waves, launch adapters, durable state, Stop integration, documentation, and tests by hangloose50

- [#20](https://github.com/Leonxlnx/unlazy/pull/20): the Windows process-tree timeout diagnosis and `taskkill /t` direction by Praveen Bishnoi

- [#24](https://github.com/Leonxlnx/unlazy/pull/24): fail-closed diagnosis of an indented `ABANDON:` by Daz Alderson

- [#25](https://github.com/Leonxlnx/unlazy/pull/25): semantic Stop-hook progress hashing by Daz Alderson

- [#26](https://github.com/Leonxlnx/unlazy/pull/26): shared-parser warning for ambiguous path-shaped EXPECT regexes by Daz Alderson

- [#29](https://github.com/Leonxlnx/unlazy/pull/29): leaf-versus-branch gate placement, visible PLAN dispatch metadata, and host-neutral skill wording by mafiaboyhacker

- [#21](https://github.com/Leonxlnx/unlazy/issues/21) and [#23](https://github.com/Leonxlnx/unlazy/issues/23): abandonment-promotion and contract-omission reports and reproducers by theislampill

- [#30](https://github.com/Leonxlnx/unlazy/issues/30): affected-Windows descriptor/path `st_dev` mismatch report and reproduction

- [#31](https://github.com/Leonxlnx/unlazy/issues/31): automatic-evidence definition drift report and false-green reproduction

Use the [skills CLI](https://github.com/vercel-labs/skills) for supported agents:

- Detailed multi-part prompts still see partial compliance and premature truncation in tested models ([Quantifying Laziness](https://arxiv.org/abs/2512.20662)).

- Reasoning can stop exploration too early or continue beyond useful compute, depending on task and model ([Thoughts Are All Over the Place](https://arxiv.org/abs/2501.18585), [When More Thinking Hurts](https://arxiv.org/abs/2604.10739), [O

- SlopCodeBench reports that no tested agent fully solved a problem end to end and that the best agent passed `14.8%` of checkpoints. Checkpoint success is not task completion ([SlopCodeBench](https://arxiv.org/abs/2603.24755)).

- s1's budget forcing lengthens reasoning by appending `Wait` multiple times when the model tries to stop; it is not a claim that one token always improves work ([s1](https://arxiv.org/abs/2501.19393)).

- METR's Time Horizon 1.1 reports a `196.5` day overall P50 doubling-time fit and `130.8` days for the post-2023 fit. The shorter figure must not be described as the all-years estimate ([METR Time Horizon 1.1](https://metr.org/blog/2026-1-2

- Closed-book knowledge-intensive tasks can hallucinate more with additional test-time compute, and compute-only post-processing cannot add ground-truth information that was not present ([Test-Time Scaling in Reasoning Models Is Not Effecti

- 2026-08-06: [Test-Time Scaling in Reasoning Models Is Not Effective for Knowledge-Intensive Tasks Yet](https://arxiv.org/abs/2509.06861), v3, COLM 2026

- 2026-07-10: [Measuring AI Ability to Complete Long Software Tasks](https://arxiv.org/abs/2503.14499), v4, NeurIPS 2025

- 2026-05-07: [SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks](https://arxiv.org/abs/2603.24755), v2

- 2026-04-13: ["Should I Give Up Now?" Investigating LLM Pitfalls in Software Engineering](https://arxiv.org/abs/2411.09916), v3

- 2026-04-12: [When More Thinking Hurts: Overthinking in LLM Test-Time Compute Scaling](https://arxiv.org/abs/2604.10739)

- 2026-01-29: [METR Time Horizon 1.1](https://metr.org/blog/2026-1-29-time-horizon-1-1/)

- 2025-12-19: [Quantifying Laziness, Decoding Suboptimality, and Context Degradation in Large Language Models](https://arxiv.org/abs/2512.20662)

- 2025-10-04: [OptimalThinkingBench: Evaluating Over and Underthinking in LLMs](https://arxiv.org/abs/2508.13141), v2

- 2025-10-03: [Context Anxiety: How AI Agents Panic About Their Perceived Context Windows](https://inkeep.com/blog/context-anxiety)

- 2025-03-01: [s1: Simple test-time scaling](https://arxiv.org/abs/2501.19393), v3

- 2025-02-18: [Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs](https://arxiv.org/abs/2501.18585), v2

- Undated page: [Unified diffs make GPT-4 Turbo 3X less lazy](https://aider.chat/docs/unified-diffs.html)

Executable file

[Current Codex releases support parallel subagents](https://developers.openai.com/codex/agent-configuration/subagents). Use the native subagent tools available in the host. When the tools are named `spawn_agent` and `wait_agent`, follow thi

[Claude Code background subagents run concurrently](https://code.claude.com/docs/en/sub-agents#run-subagents-in-foreground-or-background). Launch every leaf as a background `Agent` task, record every returned task or agent id, and seal befo

For a large regular fan-out, prefer a [Dynamic Workflow](https://code.claude.com/docs/en/workflows). Its `pipeline()` primitive runs agent work across a list under the runtime's concurrency limit. The workflow must still preserve the same s

The fenced example above is documentation. Lines inside fenced code blocks are ignored by the parser. Fence boundaries follow [CommonMark's fenced-code rules](https://spec.commonmark.org/0.30/#fenced-code-blocks): the closing marker uses th

- [Node.js child process documentation](https://nodejs.org/api/child_process.html#child_processexeccommand-options-callback)

- [Microsoft `path` documentation](https://learn.microsoft.com/en-us/windows-server/administration/windows-commands/path)

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

Executable file

Executable file

Executable file

Executable file

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

#!/usr/bin/env node

// Node 16's recursive rm rejects a directory symlink with EISDIR on some