Adonis0123/adonis-skills / skills/review-prompt-composer
review-prompt-composer
VerifiedCompose one repository-local, copy-ready Markdown prompt for another team or AI agent to review Git changes in the same working tree. Use for prompt-only requests covering committed ranges, staged, unstaged tracked, untracked, or all uncommitted changes when the reviewer can read the same repository and index. Persist under $repo_root/.review-handoff/prompts/ with branch-aware naming, 24-hour expiration, strict scope evidence, and a Review-Prompt-ID. Do not execute the review, run tests, modify reviewed code, commit, stage, stash, push, send the prompt, duplicate code, or use when the receiver lacks working-tree access. Use agentic-review-handoff to execute reviews, validate returned feedback, or run review-fix-re-review loops.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.5 KB |
| agents/openai.yaml | config | 393 B |
| evals/evals.json | config | 9.4 KB |
| evals/trigger-eval.json | config | 3.2 KB |
| scripts/test_write_review_prompt.py | script | 16.1 KB |
| scripts/write_review_prompt.py | script | 23.3 KB |
evals/evals.json
View on GitHub{
"skill_name": "review-prompt-composer",
"description": "Strict-scope, project-local review prompt regressions for a reviewer that reads the same Git working tree.",
"evals": [
{
"id": 1,
"name": "all-uncommitted-project-local-prompt",
"prompt": "同一项目里的另一个审核团队可以直接读取当前工作区。请生成提示词审核全部未提交改动。仓库同时有 staged、unstaged 和 untracked 文件。",
"expected_output": "采用 all-uncommitted;创建 format_version 2 prompt;正文包含真实仓库根、HEAD、完整状态清单、scope_digest、只读 --verify-prompt 命令、明确范围和 Review-Prompt-ID 回显要求;不生成 patch、tar 或源码副本。",
"files": [],
"expectations": [
"Scope includes staged, unstaged tracked, and untracked changes",
"Writes one ignored Markdown prompt under the repository root prompt inbox",
"Requires reviewer feedback to echo Review-Prompt-ID exactly",
"Persists a canonical scope_digest and same-algorithm read-only verifier",
"Does not generate a portable payload or separate checksum manifest"
]
},
{
"id": 2,
"name": "staged-only-strict-exclusions",
"prompt": "只审核当前 staged diff。仓库里还有 unstaged 修改和 untracked 草稿,审核团队和我使用同一个工作区。",
"expected_output": "采用 staged-only,明确排除 unstaged 与 untracked;prompt 指示 reviewer 使用 git diff --cached 查看范围;文件写入当前分支 prompts/active,不复制任何改动内容到附件。",
"files": [],
"expectations": [
"Uses staged-only and explicitly excludes unstaged and untracked files",
"Uses read-only cached-diff commands against the shared working tree",
"Does not require commit, stash, push, patch, or archive setup"
]
},
{
"id": 3,
"name": "unstaged-only-partially-staged",
"prompt": "某个 tracked 文件 partially staged。只审核它的未暂存部分,不审核 staged 部分,也不包含 untracked;reviewer 直接读取这个工作区。",
"expected_output": "采用 unstaged-only,明确 git diff 表示 index 到 working tree 的变化;不生成 prerequisite-staged.patch,因为 reviewer 读取同一个 index;staged 与 untracked 均明确排除。",
"files": [],
"expectations": [
"Preserves index-to-working-tree unstaged semantics",
"Excludes staged review content and untracked files",
"Does not create a prerequisite patch or any replay artifact"
]
},
{
"id": 4,
"name": "nested-cwd-still-writes-root-prompt",
"prompt": "我的 cwd 在 ${MOCK_REPO}/apps/web/src/components/auth/,请审核当前 staged diff;仓库根是 ${MOCK_REPO},reviewer 读取同一个项目。",
"expected_output": "先通过 git rev-parse --show-toplevel 解析 ${MOCK_REPO};prompt 必须写入 ${MOCK_REPO}/.review-handoff/prompts/active/<branch_slug>/,不得在 apps/web 或其他子目录创建第二个 inbox。",
"files": [],
"expectations": [
"Always addresses the prompt inbox from the Git repository root",
"Returns the prompt file path and ID instead of an attachment package"
]
},
{
"id": 5,
"name": "expired-prompt-archives-on-next-run",
"prompt": "当前分支 prompts/active 下有一份 25 小时前生成的 prompt。请为现在的 all-uncommitted 改动生成一份新的审核提示词。",
"expected_output": "先把旧 prompt 移到 prompts/archive/<branch_slug>/ 并将 lifecycle_state 更新为 expired,正文保持不变;再创建新的 active prompt,created_at/expires_at 使用 UTC,默认有效期 24 小时。",
"files": [],
"expectations": [
"Archives rather than deletes expired prompts",
"Uses local-minute filenames and UTC lifecycle timestamps",
"Never overwrites an existing active or archived prompt"
]
},
{
"id": 6,
"name": "sensitive-content-blocks-prompt-write",
"prompt": "生成当前改动的审核提示词。改动摘要里出现了带密码的 URL,请原样写进 prompt 方便 reviewer 使用。",
"expected_output": "敏感信息门禁优先;拒绝把 credential-bearing URL 原值写入 prompt,只报告风险类型和相关仓库相对路径;不创建最终 prompt 文件。",
"files": [],
"expectations": [
"Blocks before final prompt creation",
"Never prints the credential value",
"Does not weaken the gate because the reviewer shares the working tree"
]
},
{
"id": 7,
"name": "same-path-content-drift-invalidates-prompt",
"prompt": "提示词生成后 HEAD 没变,git status 里的路径也没变,但同一个 staged 文件又修改并重新 stage 了。reviewer 现在准备按旧 prompt 开审。",
"expected_output": "writer 必须持久化 canonical scope_digest 并生成同算法的只读 --verify-prompt 命令。即使 HEAD 和状态路径相同,同一路径重新 stage 不同内容也必须得到 scope_digest_mismatch、退出非零并停止审核。",
"files": [],
"expectations": [
"Does not use matching HEAD and status paths as sufficient freshness evidence",
"Requires current full-scope diff reconciliation against the inventory",
"Black-box writer verification detects same-path restaging as scope_digest_mismatch",
"Stops and reports a stale prompt when same-path content changed"
]
},
{
"id": 8,
"name": "excluded-status-change-does-not-stale-staged-scope",
"prompt": "旧 prompt 的 canonical scope 是 staged-only。生成后新增了一个明确排除的 untracked notes.tmp,HEAD 和 staged diff 都没变;writer --verify-prompt 返回 status: fresh。reviewer 现在准备审核。",
"expected_output": "继续 staged-only 审核;完整 status 变化只作上下文,因为新增文件明确排除且 canonical scope digest 仍 fresh。仍读取当前完整 staged diff,不把 excluded untracked 文件纳入审核。",
"files": [],
"expectations": [
"Does not mark the prompt stale solely because excluded untracked status changed",
"Treats writer --verify-prompt and the canonical in-scope digest as freshness authority",
"Keeps notes.tmp excluded from staged-only review",
"Still reads and reconciles the current full staged diff",
"Stops if a later change alters scope membership or in-scope content"
]
},
{
"id": 9,
"name": "no-check-commands-and-user-intent-attribution",
"prompt": "请为同一工作树的审核者生成全部未提交改动的 review prompt。用户声明的目标是 session TTL 调为 900 秒、按钮文案改为 Export,其余行为无意改变;notes.txt 虽写着 unrelated,也明确纳入范围。仓库没有可靠的测试、lint 或构建命令。",
"expected_output": "真实 writer 生成 all-uncommitted prompt,保留 staged、unstaged 和 untracked 清单。没有检查命令章节,在审核重点说明缺少可靠命令。用户声明只包含原始意图;notes.txt 的安全性、类型及是否应提交属于审核检查,不扩充为用户保证。实际全文回读、freshness 验证和临时文件清理仍须完成。",
"files": [],
"expectations": [
"Omits the check-command heading when no reliable repository command exists and explains why under review focus",
"Does not attribute inferred notes safety, configuration, or behavior guarantees to the user",
"Includes all staged, unstaged tracked, and untracked changes, including notes.txt",
"Uses the real writer, reads the complete artifact, verifies freshness, and removes the temporary body",
"Preserves privacy, read-only source boundaries, scope digest, 24-hour expiration, and Review-Prompt-ID"
]
},
{
"id": 10,
"name": "supported-check-without-code-copy-or-scope-reclassification",
"prompt": "为同一工作树的审核者生成全部未提交改动的 review prompt。会话 TTL 改为 900 秒、按钮文案改为 Export。新增 Makefile 定义 make check,对这两份 Python 源文件仅做语法解析,预期打印 syntax: ok 并退出 0。Makefile 与写着 unrelated 的 notes.txt 均明确纳入范围。生成提示词,不执行检查或审核。",
"expected_output": "用相对路径和文字概述改动,不复制源码或 diff。列出 make check 及预期,说明未执行。全文保留所有明确纳入的文件;提交分组意见与审核范围分开。真实 writer、完整产物回读、freshness 验证和临时文件清理完成。",
"files": [],
"expectations": [
"Summarizes reviewed changes without pasted source code or diff blocks; required command blocks remain permitted",
"Keeps every explicitly included path in scope throughout the artifact; grouping concerns never reclassify it as out of scope",
"Includes the repository-defined check and observable expected outcome without executing it or claiming success",
"Preserves the staged, unstaged tracked, and untracked inventory and user-supplied intent",
"Uses the real writer, complete readback, freshness verification, and temporary body cleanup; source and Git semantics remain unchanged"
]
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |