菜单
Compose a MissionBrief YAML block to prepend to any Agent dispatch during the team-ownership dev-contract pilot (2026-05-23 → 2026-06-15). Use BEFORE spawning kernel-warden or any future warden. Fails if missing end-state, why, authority-bounds, or success-criteria.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Agentic harness diagnostic + improvement loop. Probes the framework with real model runs, uses the rax-diagnose CLI to root-cause failures from structured trace data, ships ONE coordinated architectural fix, verifies via before/after diff, commits with empirical evidence. Use at the start of any harness improvement session — replaces ad-hoc grep + log-spelunking with a deterministic feedback loop.
Use when the architecture may have drifted from documentation, packages have grown complex, dead code or disabled systems are suspected, or before planning a major refactor — scoped to the reactive-agents-ts 22-package monorepo.
Reactive Agents framework architecture — layer stack, dependency graph, build order, and package structure. Use when planning work, understanding package relationships, or determining build dependencies.
Use when analyzing the Reactive Agents codebase for architectural improvements, abstraction opportunities, composability gaps, or Effect-TS engineering quality — before proposing refactors, during design reviews, or when codebase complexity is growing.
LLMService API contract — the correct signatures for complete(), stream(), embed(), and response types. Use when calling LLMService from any layer, writing reasoning strategies, or building LLM-dependent features.
LLM provider streaming patterns and per-provider quirks. Use when adding a new provider, implementing adapter hooks, or debugging streaming tool call behavior in packages/llm-provider.
| name | mission-brief |
| description | Compose a MissionBrief YAML block to prepend to any Agent dispatch during the team-ownership dev-contract pilot (2026-05-23 → 2026-06-15). Use BEFORE spawning kernel-warden or any future warden. Fails if missing end-state, why, authority-bounds, or success-criteria. |
Mandatory input contract for warden dispatch during the team-ownership pilot.
Agent call targeting a warden agent (kernel-warden, future wardens).Prepend this YAML block (inside a fenced code block, language yaml) at the very top of the Agent prompt:
mission-brief:
end-state: <one sentence, measurable. e.g., "Loop-detector streak counter resets on ACTION steps only; existing 2458 tests still pass.">
why: <one sentence, the constraint or motivation. e.g., "Observations resetting the streak masked IC-1 loop bug; project memory entry feedback_clean_types.md.">
key-tasks: <ordered list, ≤5 items>
authority-bounds:
paths: <glob list of paths the warden may Read/Edit>
commands: <list of bash commands the warden may run>
out-of-scope: <explicit list — refuse rather than escalate silently>
success-criteria: <deterministic gates: tests-green | file:line landmark | LOC delta ceiling | ablation-warden lift threshold>
retries-allowed: <integer, default 2>
Refuse to dispatch if any required field (end-state, why, authority-bounds.paths, success-criteria) is missing or contains "TBD" / "TODO" / vague phrasing.
mission-brief:
end-state: "loop-detector.ts:102 streak counter resets only on ACTION steps; 2458 tests green."
why: "IC-1 (Apr 12) regression — observations were resetting streak, masking infinite loops."
key-tasks:
- Read loop-detector.ts to locate streak-reset call sites
- Write failing test asserting streak persists across observation
- Implement guard
- Run typecheck + targeted test
- Run full reasoning package suite
authority-bounds:
paths:
- packages/reasoning/src/kernel/capabilities/reflect/loop-detector.ts
- packages/reasoning/src/kernel/capabilities/reflect/*.test.ts
commands:
- bunx turbo run typecheck --filter=@reactive-agents/reasoning
- bun test packages/reasoning/src/kernel/capabilities/reflect/
out-of-scope:
- Edits outside packages/reasoning/src/kernel/
- Changes to act.ts or terminate.ts
success-criteria: "Targeted test fails before, passes after; full reasoning suite green; no LOC delta > 30."
retries-allowed: 2
This skill is part of the team-ownership dev-contract pilot. See [[2026-05-23-team-ownership-dev-contract-pilot]]. If pilot is killed on 2026-06-15, this file is removed in the revert commit.