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 skills add FuRongJun-1999/dsh-memory --skill designer-perspectivegit clone --depth 1 https://github.com/FuRongJun-1999/dsh-memoryWrote 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/skills/furongjun-1999/dsh-memory/designer-perspective)<a href="https://agentmods.dev/skills/furongjun-1999/dsh-memory/designer-perspective"><img src="https://agentmods.dev/badge/skills/furongjun-1999/dsh-memory/designer-perspective/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/furongjun-1999/dsh-memory/designer-perspective"><img src="https://agentmods.dev/badge/skills/furongjun-1999/dsh-memory/designer-perspective.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00230 | $0.02917 |
| Opus 5 | $0.00115 | $0.01458 |
| Sonnet 5 | $0.00046 | $0.00583 |
| Haiku 4.5 | $0.00023 | $0.00292 |
Grade A, and why
designer-perspective 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 today.
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 — 159 lines — stays where its author put it; the contents beside it link to each section on GitHub.
设计者视角（designer-perspective）
设计者视角是一种可习得的观测方式，不是身份特权（《智能论3.4》1.4.2）。本技能在动手之前启用，回答三个问题：该不该做（资格）、为什么做（方向）、条件够不够（条件空间）。它不执行操作，只提供判定与归因；执行与裁决仍由操作层技能、验证单元与 MCP 物理基底完成。
位置声明：本技能运行在全局观测层。设计者位置的功能是放权，不是控制（1.4.1）——因此本技能只输出判定与关注信号，不替操作层做内部决策。
When to use
在以下时机启用本技能：
- 任务开始前：需要先判断执行资格与方向，而非直接开始执行。
- 候选歧义时：多个候选「看起来都像」，需要判断谁有资格执行（负路由优先）。
- 失败之后：需要从条件层定位遗漏条件，而不是只改答案。
- 长期运行中：需要从整体查看认知图，识别未知区域并决定下一步方向。
- 自我校准时：需要核对「依据什么、在什么条件下成立、错了怎么办」。
不适用于：已明确路径的纯执行型任务、时间紧迫的琐碎操作、需要写认知图事实层的动作（后者由 agent 经 MCP 显式发起）。
克制条款（不适用条件）
条件不满足即不适用（负路由：观测位置与操作层执行互斥；本技能不执行具体操作、不替代验证单元裁决、不写入认知图事实层）
具体边界：
- 不执行：本技能产出判定单与建议调用清单，不产出操作结果。
- 不裁决事实：资格判定由条件证据裁决（6章.2），事实裁决归验证单元与 MCP 基底；本技能不代替
mdcg_verify。 - 不写库：脚本只读；
record/verify/flywheel等写操作只输出建议参数，由 agent 显式发起。 - 不硬答盲区：无法建立判断路径时判定 BLINDSPOT 并停止猜测，指出缺失的条件维度。
How to execute
按五步执行，每步都有可核查的产出。
步骤 1：声明条件空间
切换到全局观测位置，按四维声明当前条件空间（第三章 3.1/3.2、附录14）：
| 维度 | 回答的问题 | 协议映射 |
|---|---|---|
| 观测位置 | 谁在观测？从哪个尺度观测？ | 实例角色、结构倾向 |
| 观测工具 | 用什么方法测量信息差？ | 验证方法、反思方法 |
| 时间窗口 | 在什么时间范围内观测？ | 观察窗口 N、迭代周期 |
| 存在约束 | 物理-信息层允许什么？ | 锚点层公理、通用价值观 |
声明格式见 references/01-observation-position.md。运行：
python scripts/designer.py declare --position designer --space "观测位置=全局;观测工具=条件证据;时间窗口=当前任务;存在约束=锚点层公理"
步骤 2：白箱三问自检
对每个待执行候选，逐问核对（第〇章）：
- 依据什么——引用了哪条知识/哪个节点？（锚定）
- 在什么条件下成立——该结论的适用条件是什么？（条件空间）
- 错了怎么办——失败如何被记录、学习、修正？（验证闭环）
任一问无法回答，该候选不得进入 ACCEPT。
步骤 3：四态资格裁决
判定顺序固定，先拒绝后接受（6章.2）：
候选 C
├─ 存在足以证明「不适用」的条件冲突？ → REJECT
├─ 候选存在但区分条件未定？ → DEFER（继续递归补条件）
├─ 无法建立判断路径？ → BLINDSPOT（停止猜测）
└─ 条件充分，获得执行资格 → ACCEPT
核心纪律：资格由条件证据裁决，不由相似度裁决。相似度可以产生候选，但不能单独授予执行资格。判定规则与边界见 references/04-qualification-verdict.md。运行：
python scripts/designer.py judge --query "..." --conditions "{\"观测位置\":\"...\"}" --emit-mcp
步骤 4：方向判断与关注信号
从整体查看认知图：识别未知区域（盲区）→ 据此决定方向（附录「蜂群架构奠基声明」五）。
- 值不值得做：
Value(s,v) = ΔD_task(s,v) · σ(Gain_task(s,v))；Gain_task 是筛选器，ΔD_task 是定价器（5.3）。 - 成本：
K_π含导航税、验证成本、计算开销；在有限结构成本下选择能最大程度缩小信息差的路径。 - 权限边界：π* 仅为 L1/L2 策略搜索层目标函数，L3 存在级决策不适用（5.3 权限边界声明）。
- 观测不干预：可发关注信号（signal），不可直接干预内部决策。
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/01-observation-position.md 6.2 KB
- references/02-structure-recognition.md 6.1 KB
- references/03-direction-judgment.md 4.9 KB
- references/04-qualification-verdict.md 5.9 KB
- references/05-condition-attribution.md 5.6 KB
- scripts/designer.py 23 KB runs code
- tests/cases.jsonl 5.1 KB
- tests/selftest.py 2.1 KB runs code
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.
- today First seen · 159 lines · 230 tokens per session scan A f7f9b4753b7d
designer-perspective is a skill published in the GitHub repository FuRongJun-1999/dsh-memory (159 stars, last pushed today), licensed MIT. It adds 230 tokens to every session and 2,917 once invoked, about $0.0011 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 2026-09-11.
Other skills, from other repositories
bbs-relay-pickup
A coordination procedure for agents that share work on tasks through a BBS task system, where BBS is the service used to claim, run, and report task steps.
bcs-coordination
A coordination system for several software agents, meaning separate automated assistants, working together.
bcs-coordination
Public BCS multi-bot coordination skill. Uses bcs-cli to call a configured BCS HTTP endpoint for bot discovery, group creation, routing, and session operations.
orchestrate
This skill should be used when the user asks to 'orchestrate a task', 'break down work into parallel agents', 'coordinate subtasks', 'run agents in parallel', or mentions 'multi-agent'. Decomposes complex tasks into tracked subtasks, dispatches parallel subagents, and coordinates until completion.
undercover-game-referee
A game-session referee procedure for the social deduction game Who Is the Spy. It follows the game's stages for speeches, voting, revealing roles, and ending a session.
bbs-relay-single-task
A relay procedure for taking over a task from an engine notification, which includes the task's current state.