跳到正文
cccskills
免费在 GitHub 公开

prohibit-bad-habits

Use when a user wants to define patterns the agent should not do. Don't use for remote, credential, publish, deploy, or irreversible changes.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md5.5 KB
  • agents/openai.yaml138 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Prohibit bad habits

Contract

FieldBound contract
TriggerHuman requests that specific agent behavior patterns be documented as prohibited in a structured project document or agent-rule file rather than in AGENTS.md.
AuthorityReversible local: writes only the named local artifact; rollback is deleting the written file if it did not exist before and the human has not confirmed it. No remote mutation.
Side effectLocal file write: creates or updates one structured project document or agent-rule file that enumerates prohibited agent behavior patterns.
DoneA structured artifact exists at the agreed target path and contains at least one prohibited-habit entry with a non-empty name and description, verified by the human.

Inputs

  • Prohibited patterns: required. For each pattern, the human supplies a short name, a concrete description of the behavior to avoid, the negative consequence of that behavior, and optionally a category and a workaround or alternative. If the human cannot supply a concrete description, stop rather than invent one.
  • Target location: optional. The human may specify the exact file path. Default: docs/prohibited-habits.md for shared project docs; agent-rules/prohibited-habits.yaml for agent-rule locations.
  • Authority confirmation: the human confirms or revises the target path before the write occurs.

Procedure

  1. Gather prohibited patterns. Ask the human to list each pattern to prohibit. For each, collect: name (short identifier), description (concrete behavior to avoid), consequence (what goes wrong when the agent does this), and optionally category and workaround. If the human cannot supply a concrete description, stop without writing. Done when: every pattern the human listed has a name and a concrete description, and any pattern lacking a description is named as blocked rather than invented.

  2. Confirm target location. Accept the human's specified path, or use docs/prohibited-habits.md for a shared-project location or agent-rules/prohibited-habits.yaml for an agent-rules location. Present the chosen path to the human and wait for confirmation before proceeding. Done when: the human has confirmed or revised the path, and the path is recorded for the write.

  3. Choose format. Use Markdown (*.md) if the target is under docs/ or the human requests it. Use YAML (*.yaml) if the target is under agent-rules/ or multiple agent-rule files already exist in the project. Otherwise default to Markdown. Done when: the format matches the target path extension, or the default is chosen and the path extension agrees with it.

  4. Draft file content. Write the following structure to the confirmed path:

    • Markdown format:
      # Prohibited Habits
      
      ## <Pattern Name>
      
      **What to avoid:** <concrete description of the behavior>
      **When it happens:** <negative consequence of the behavior>
      **Workaround:** <alternative or mitigation, if provided>
      
    • YAML format:
      prohibited_habits:
        - name: <short name>
          description: <concrete behavior to avoid>
          consequence: <what goes wrong>
          category: <category, if provided>
          workaround: <alternative or mitigation, if provided>
      

    Done when: the drafted content contains one entry per gathered pattern, each entry carries name, description, and consequence, and the structure matches the chosen format.

  5. Write the file. Create or overwrite the confirmed target path with the drafted content. If a pre-existing file at that path contains content the human has not reviewed in this session, stop and report the conflict rather than overwriting silently. Done when: the file exists at the confirmed path and its contents match the draft, or the pre-existing-conflict stop condition fired and no write occurred.

  6. Present to the human. Show the written file path and content. The human confirms acceptance or requests revision. If the human declines to review, report the partial result and stop without claiming done. Done when: the human has confirmed acceptance of the written content, or the human declined to review and the result is recorded as partial.

Failure and recovery

  • Non-concrete input: if the human cannot describe a pattern concretely, stop and return blocked: non-concrete-description. Do not substitute a generic warning.
  • Unreviewed pre-existing file conflict: if the target file exists with unreviewed content, stop and return blocked: unreviewed-file-exists. Do not overwrite.
  • Human declines review: return partial: file-written with the file path. The done predicate does not hold until the human confirms acceptance.
  • Rollback: delete the written file only if (a) the file did not exist before this session and (b) the human has not confirmed the content. If the file existed before, restore it from the original content if possible. Rollback does not apply after human confirmation.

Output

  • done: prohibited-habits-documented with the file path. The human confirmed the written artifact. - partial: file-written with the file path. The file was written but the human declined to review. - blocked: non-concrete-description: the human could not describe a prohibited pattern concretely. - blocked: unreviewed-file-exists: a pre-existing file at the target path was not reviewed in this session.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Use when explaining System V AMD64, ARM AAPCS, RISC-V psABI, stack frames, variadic calls, or FFI register rules. Not for the Rust FFI binding layer: use rust-ffi.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

Use when configuring ADC sampling time, DMA-driven ADC, calibration, or DAC channel setup on bare-metal MCUs. Not for the DMA stream itself: use dma-baremetal.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

af-xdp

免费

Use when creating AF_XDP sockets, configuring UMEM and XSK rings, writing an XDP redirect program, or choosing copy versus zero-copy mode. Not for full kernel bypass: use dpdk.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

Use when a completed session needs an agent-environment retrospective. Not for an engineering retrospective from telemetry: use engineering-retrospective.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

Use when a redacted, trimmed agent transcript must be appended to a GitHub PR or issue body, with human approval and preview. Not for automated or model-initiated insertion.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

agents-md

免费

Use when a repo needs agent setup, AGENTS.md added or made lean, CLAUDE.md audited, or agent instructions scored or pruned. Not for remote, credential, publish, deploy, or irreversible changes.

中文简介正在准备中。当前显示原文说明。

OutlineDriven/odin-claude-plugin382026年9月29日 更新

查看 OutlineDriven 的全部技能

报告此技能的问题