Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/bladeofgod/flutter-ai-harness/execute-tasksgit clone --depth 1 https://github.com/bladeofgod/flutter-ai-harnessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/bladeofgod/flutter-ai-harness/execute-tasks)<a href="https://agentmods.dev/commands/bladeofgod/flutter-ai-harness/execute-tasks"><img src="https://agentmods.dev/badge/commands/bladeofgod/flutter-ai-harness/execute-tasks.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00019 | $0.02453 |
| Opus 5 | $0.00010 | $0.01226 |
| Sonnet 5 | $0.00004 | $0.00491 |
| Haiku 4.5 | $0.00002 | $0.00245 |
Grade A, and why
execute-tasks scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
执行 $ARGUMENTS 中的任务卡。只有实现、验证、Review、修复和复审全部完成,任务卡才算完成。
前置检查
- 对每个输入路径先做词法规范化,并确认规范化后仍是仓库
docs/tasks/直属目录下、文件名为 lowercase kebab-case 的普通非符号链接 Markdown 文件;从仓库根到任务文件的每个路径组件 都必须是非符号链接的普通目录或文件,解析后的真实路径也必须位于仓库内。拒绝..逃逸、done/、子目录、符号链接、特殊节点和仓库外路径,再读取任务内容。 - 编辑前完整阅读所有卡片。
- 按声明依赖排序,再按文件名排序。
- 阻塞项未解决或外部依赖缺失时停止。
- 保护无关工作树改动,并将 diff 限定在当前卡范围。
Executor 分流
- 缺少
executor:停止并报告无效任务卡,不猜测执行角色。 - 缺少
platforms或workKinds的活动任务:停止并报告无效任务卡;归档历史任务按 Harness 兼容规则处理,不重新执行。 task-executor:只执行 Dart Client、Flutter、文档、规划、Harness、传输中立 Capability Contract 及其允许的单 Runtime 工作。android-engineer:只执行platforms: [android]的 Native Module、Bridge Adapter 或 Android 单平台门禁,并加载 Kotlin/Android 与原生测试 Skill。ios-engineer:只执行platforms: [ios]的对应工作,并加载 Swift/iOS 与原生测试 Skill。bridge-engineer:只执行结构化 Wire Contract 或包含多个 Runtime 的最终集成,要求契约和 所有声明平台证据齐全。- 其他值:停止并报告无效任务卡。
只依据 frontmatter 选择 Agent 并校验结构化范围,不根据文件名、正文或 diff 猜测平台。 Executor/范围冲突或跨平台实现未拆卡时停止;正文和实际 diff 是否诚实匹配声明范围由 Reviewer 检查。
普通 Review Profile 分流
普通 Review 只由任务完整 workKinds 确定,不增加任务字段,也不根据文件名、扩展名或 diff 猜测:
flutter、dart-client、native、bridge-adapter、integration、quality-gate选择code-reviewer。harness选择harness-reviewer。capability-contract、bridge-contract选择contract-reviewer。documentation、planning只有在没有命中上述实质 Profile 时才选择contract-reviewer;作为实现 配套工作时不额外增加 Profile。
结果按 code-reviewer、harness-reviewer、contract-reviewer 固定顺序去重。一个任务可以命中多个
Profile。security-reviewer 是独立安全维度,不参与普通 Profile 映射。
单卡闭环
- 严格按卡片范围实现代码和测试。
- 格式化触碰的 Dart 文件。
- 通过
scripts/quality/capture-evidence.sh直接执行受影响静态分析、聚焦测试和make lint,把 shell-safe 命令、工具版本、退出码、稳定结果或首个失败根因、完整脱敏输出行数/字节数与 SHA-256 的bounded-v1摘要写入docs/reviews/test-evidence/<task-slug>.log。每个正式自动化 Gate 使用稳定的 canonical lowercase kebab-case ID,通过--gate <gate-id>更新;同一 Gate 只保留被后续尝试取代的 首次失败和最新结果,新 Gate 按首次出现顺序增加。命令只执行一次,原始输出只进入临时文件并在脱敏后删除。本地执行不伪造 Artifact;CI 通过--artifact把同次命令的完整脱敏输出交给固定 14 天上传步骤。不得直接重定向 stdout/stderr 到入库证据或 Artifact。 - Harness 实现或修复轮使用
make harness-test-focus显式声明受影响的HARNESS_FIXTURE_CASES或HARNESS_FIXTURE_PATHS;Evidence 必须记录选择条件、实际 case 数和 ID。选择器缺失、非法、未知或零匹配 必须失败,聚焦 Gate 不得宣称未执行的 Fixture 已验证。新增 Harness 规则默认在所属领域 Dart 测试中从 明确合法 Fixture base 创建结构化 mutation;只有改变既有 Legacy Shell 场景或完整性迁移时才重采集 inventory/catalog。影响面无法可靠收窄,或修改共享 Entity、公共包 API、协议生成、DI 装配、路由或平台 契约时升级验证范围。 - 首轮实现和自动化验证结束后冻结候选,在全部适用普通 Profile 与条件性 Security Review 返回前不得
修改实现、测试或任务范围。运行环境支持时并行启动独立审查;不支持时可顺序调用,但必须绑定同一候选,
且不得把任一 Reviewer 结论提供给其他 Reviewer。把普通 Profile 结论聚合写入唯一的
docs/reviews/execute-<task-slug>.md:frontmatter 必须包含与任务文件 basename 一致的taskslug、status、当前未解决总数p0、p1、reviewFormat: routed-v1,以及与映射结果完全相同的有序reviewProfiles。正文按 Profile 分节,每条发现标出ownerProfile;任一 Profile 仍有 P0/P1 时聚合 报告不得通过。 - 判断是否需要 Security Review。任务声明
securityReview: required时必须执行;未声明但实际 diff 引入或改变下列任一边界时,先把该字段补入活动任务卡再执行:- 认证、会话、授权、用户数据隔离、凭据、隐私数据或安全存储。
- 网络、文件、Deep Link、WebView、不可信输入反序列化或其他攻击者可控输入。
- MethodChannel/EventChannel、原生权限、Manifest、Entitlements 或平台安全配置。
- 第三方依赖、GitHub Action、构建/安装/生成脚本或依赖来源。
.claude/settings.json、MCP、CI 权限,或 Command/Agent/Skill、适配生成器与脚本中会改变读写、命令、网络、凭据、提交或发布能力的执行语义;纯描述修正和未改变能力的生成适配同步不触发。
- 需要时由调用工作流运行必要的只读验证,再调用没有 Bash/写入能力的
security-reviewer;写入docs/reviews/security-<task-slug>.md时列出实际审查的 implementationFiles,并在所属任务归档前用implementation_digest.dart为最终实现生成 implementationDigest。其首轮必须与所有普通 Profile 保持独立,不读取普通 Review 结论;不需要时不生成跳过报告。任务归档前任一 implementationFiles 内容变化都必须重新审查并更新摘要。 - 全部首轮结论返回后一次性汇总 P0/P1,再使用
fix-review-findings进行一个修复批次。每轮只重跑修复 实际影响的 Gate、发现所属普通 Profile、审查输入或语义被改变的其他普通 Profile,以及implementationFiles内容变化或新增/改变安全边界时的 Security Review。修复轮只更新实际重跑的 Gate,不为未执行命令制造记录;Evidence 的 Gate 更新、报告状态更新 或与结论无关的生成适配同步本身不使全部 Profile 失效;任务范围、验收标准或权威契约变化时,使所有 受影响 Profile 失效。把每轮 Gate/Profile/Security 失效判断及理由记录在聚合报告中。execute-tasks已包含实现授权;自动修复最多三轮,超过后停止并请求用户决策。 - Harness 任务的全部 P0/P1、聚焦 Gate 和适用复审清零后,对最终准备归档的候选执行一次无过滤的
make harness-test并记录 full Harness Gate;聚焦 Gate 不能替代它。完整回归失败时只修复实际失败区域并 重新执行受影响 Gate/Profile,形成新的最终候选后再运行完整回归。CI、make check和发版检查始终全量。 - 将完成任务卡移入
docs/tasks/done/,并更新其他仓库内任务引用;Review、Security Review、Evidence、 UI Spec、Audit 和 App Operator 报告都不得随任务移动或重写。 - 归档完成后重新运行
make harness-check;归档任务缺少已完成依赖、通过的普通 Review、要求的 Security Review、合法的安全快照元数据或测试证据时门禁必须失败。已归档报告不要求其摘要匹配后来 工作树内容;后续修改由新的任务、Review、Evidence 和摘要负责。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 115 lines · 19 tokens per session scan A 061461e3f612
execute-tasks is a command published in the GitHub repository bladeofgod/flutter-ai-harness (113 stars, last pushed 22d ago), licensed MIT. It adds 19 tokens to every session and 2,453 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
git
Git operations with intelligent commit messages and workflow optimization.
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.