chore: drop another stray scratch file

This commit is contained in:
Huanqi Cao
2026-08-03 22:50:25 +08:00
parent 9ae358e087
commit b43358e215

View File

@@ -1 +0,0 @@
[{"comments":{"nodes":[{"body":"🟡 **warning**: Duplicated contract with a stale example import: the §Managed shell environment section carries the full bashEnv registry contract (both prose paragraphs, the contributor code example, and the overlay paragraph) near-verbatim duplicated in this PR's new owning home, packages/bash/bash-env/README.md. The tool-bash copy's example still reads `import type {} from '@deepseek-ai/dsh-tool-bash'` to get `ctx.bashEnv` typed — but this PR moved the `declare module 'cordis'` merge to dsh-bash-env (tool-bash only type-imports it, so the example compiles transitively but names the wrong owner). tool-pwsh's README shows the intended post-extraction shape: a one-paragraph summary linking to ../bash-env/. Trim tool-bash's section to the bash-specific facts plus the link, keep the contract and example in bash-env only (one home per fact; symmetry between the two shell tools), update both language sides and re-record the pairing. Non-blocking, but worth fixing before merge since the stale example is a direct consequence of this PR's extraction.\n\n\u003csub\u003e🤖 v5\u003c/sub\u003e"}]},"id":"PRRT_kwDOS3Pfcs6V_H8o","isResolved":false,"line":30,"path":"packages/bash/tool-bash/README.md"},{"comments":{"nodes":[{"body":"🔵 **suggestion**: Nit: `Scenario.pwshOnly`'s JSDoc says the composition \"needs a real `pwsh` on PATH\", but the probe is caller-owned and the shipped caller (examples/acp-agent/tests/acp.snapshot.ts:441) deliberately follows resolvePwshPath() — Program Files installs are found even when bare `pwsh` is not on PATH. This is the same stale \"on PATH\" phrasing already corrected in the executor.spec.ts and integration.spec.ts headers this round. One-word fix: \"a usable `pwsh`\".\n\n\u003csub\u003e🤖 v5\u003c/sub\u003e"}]},"id":"PRRT_kwDOS3Pfcs6V_H8w","isResolved":false,"line":164,"path":"packages/support/acp-snapshot/src/suite.ts"},{"comments":{"nodes":[{"body":"🔵 **suggestion**: Optional nit: the `# task_output/task_kill stay mounted so background pwsh runs are readable and killable.` contract comment sits at column 0 mid-way through the acp-agent config map (between `skills` and `goals` keys), in both pwsh.cordis.yml and pwsh.cordis.snapshot.yml. Valid YAML and a genuinely useful pin (the mounting decision is the bundle's toolTasks default, invisible in this file), but indent it into the block or move it above the composition's plugin list where the decision reads naturally.\n\n\u003csub\u003e🤖 v5\u003c/sub\u003e"}]},"id":"PRRT_kwDOS3Pfcs6V_H82","isResolved":false,"line":30,"path":"examples/acp-agent/tests/pwsh.cordis.yml"},{"comments":{"nodes":[{"body":"🟡 **warning**: 拒绝 shell 工具与执行器方言不匹配: 当组合误将 `tool-pwsh` 与 `dsh-bash-local` 搭配时当前注入会正常完成PowerShell 命令却被交给 `bash -c`,通常只返回普通的非零退出标记;反向将 `tool-bash` 搭配 `pwsh-local` 也一样。新增第二种方言后,`ctx.bash` 需要暴露可校验的方言能力,并让两个工具在加载时拒绝不匹配的执行器,否则明显的部署错误会静默表现为命令失败。\n\n\u003csub\u003e🤖 v6\u003c/sub\u003e"}]},"id":"PRRT_kwDOS3Pfcs6V_H88","isResolved":false,"line":141,"path":"packages/bash/tool-pwsh/src/index.ts"}]