跳到正文
cccskills
免费在 GitHub 公开

claudeception

Use when reviewing what was learned after debugging, workaround discovery, or trial-and-error investigation, and capture it as a dated journal entry under ~/.codex/journal

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(10)

  • SKILL.md4.4 KB
  • examples/2026-01-15-nextjs-server-side-error-debugging.md760 B
  • examples/2026-01-16-prisma-connection-pool-exhaustion.md806 B
  • examples/2026-01-17-typescript-circular-dependency.md740 B
  • LICENSE1.0 KB
  • README.md4.0 KB
  • resources/journal-entry-template.md718 B
  • resources/research-references.md7.5 KB
  • scripts/claudeception-activator.sh1.8 KB
  • WARP.md1.3 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Claudeception

Claudeception captures episodic memory from work sessions.

It does not create or update skills. Its job is to preserve what happened, what was learned, and why it mattered in a dated journal entry that can be searched later or promoted through a separate workflow.

When To Use

Use this skill when any of these are true:

  • A task required non-obvious debugging or investigation
  • A workaround or recovery path was discovered through trial and error
  • An error message was misleading and the real root cause was learned
  • A project-specific or system-specific constraint was uncovered
  • The user asks /claudeception
  • The user says "save this as a skill", "extract a skill from this", or "what did we learn?"

Do not use this skill to create durable procedural knowledge. If the learning should become a reusable skill, leave that for a separate promotion step.

Core Rule

Never create a skill from this skill.

Even when the user says "save this as a skill", honor the intent by writing a journal entry unless they explicitly ask for a separate skill-writing workflow.

Capture Threshold

Write a journal entry when the session produced knowledge that is worth preserving, even if it is narrow or specific to one system.

Good journal material:

  • Incident-specific discoveries
  • System quirks and deployment gotchas
  • Root causes behind confusing failures
  • Investigated tradeoffs and workarounds
  • Repeated patterns that are not generalized yet

Skip capture when:

  • The task was a routine documentation lookup
  • The conclusion is already covered cleanly by an existing skill or project docs
  • The note would mostly restate obvious knowledge
  • The note would require storing secrets or sensitive internal details

Output Location

Write entries to:

~/.codex/journal/YYYY-MM-DD-topic-slug.md

Examples:

  • ~/.codex/journal/2026-04-05-fastly-draft-version-publish-mismatch.md
  • ~/.codex/journal/2026-04-05-nextjs-server-side-error-debugging.md

If multiple notes land on the same date with the same slug, add a short suffix.

Entry Format

Use the template in resources/journal-entry-template.md.

Every entry should include:

  • Frontmatter with date, title, tags, and source_trigger
  • A short factual description of what happened
  • The actual learning, not just the symptoms
  • Evidence such as errors, commands, files, or observations
  • A brief reuse signal describing whether this looks one-off or recurring

Keep the note concrete. Journal entries are not polished runbooks.

Workflow

  1. Review the task or session
  2. Isolate the single learning worth preserving
  3. Choose a dated filename with a concise slug
  4. Write the journal entry using the template
  5. Include exact evidence and the real takeaway
  6. Stop after writing the entry

Do not create a new skill. Do not update an existing skill. Do not generalize beyond what the evidence supports.

Retrospective Mode

When /claudeception is invoked at the end of a session:

  1. Review the session for journal-worthy learnings
  2. Prefer one focused entry per distinct incident or discovery
  3. If there are multiple unrelated learnings, write separate entries
  4. Summarize what was captured and where it was written

Quality Gates

Before finalizing a journal entry, verify:

  • The filename is dated
  • The note describes something actually observed or verified
  • The writeup includes evidence or concrete trigger conditions
  • The note preserves local context instead of inflating it into a general skill
  • No secrets, credentials, or sensitive URLs are included
  • No new skill files were created

Anti-Patterns

  • Turning every discovery into a reusable skill
  • Rewriting the note as a polished runbook
  • Stripping away the concrete system context that made the discovery useful
  • Capturing vague conclusions without evidence
  • Creating duplicate journal entries for the same learning in the same session

Self-Check Prompts

Ask:

  • What happened that was not obvious at the start?
  • What was the actual root cause or constraint?
  • What would future-me want to grep for?
  • Is this a concrete incident note or a reusable procedure?

If it is an incident note, write it to the journal.

If it is a reusable procedure, do not create it here. Leave it for a separate promotion workflow.

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

Fix ArgoCD ExternalSecret deployment failing with "namespace X is not permitted in project Y". Use when: (1) ExternalSecret shows OutOfSync in ArgoCD but won't sync, (2) ArgoCD application status shows "namespace X is not permitted in project 'infrastructure'", (3) ExternalSecret targets a namespace managed by a different ArgoCD project, (4) Using apps-of-apps pattern with separate infrastructure and application projects.

中文简介正在准备中。当前显示原文说明。

divinevideo/divine-mobile2662026年10月19日 更新

Art direction for any content — reads text, PDF, Word, HTML, PPT, then proposes 2-3 creative directions with photography style, mood, and visual language. After selection, generates AI image prompts and visual briefs section-by-section. Use when the user shares content and needs visual direction, image sourcing, or creative direction for any material.

中文简介正在准备中。当前显示原文说明。

divinevideo/divine-mobile2662026年10月19日 更新

Fix "Null check operator used on a null value" errors when an object is set to null during an async await. Use when: (1) Object reference is nullified while awaiting, (2) Code accesses object with ! after await returns, (3) Cancel/dispose operations run concurrently with async operations on same object. Solution: capture local reference before await.

中文简介正在准备中。当前显示原文说明。

divinevideo/divine-mobile2662026年10月19日 更新

Add custom metadata headers (x-amz-meta-*) to AWS v4 signed requests for GCS S3-compatible API. Use when: (1) Adding custom metadata to GCS uploads via S3 API, (2) Getting signature mismatch errors after adding new headers, (3) x-amz-meta-* headers being ignored or causing 403 errors. Custom headers MUST be included in canonical headers and signed headers list.

中文简介正在准备中。当前显示原文说明。

divinevideo/divine-mobile2662026年10月19日 更新

Fix password/secret authentication failures caused by trailing newlines when creating Google Cloud secrets (or similar) with bash here-strings. Use when: (1) Password authentication fails with correct password, (2) Secret created with `<<< "value"` syntax, (3) Error like "password authentication failed" or "invalid token" despite correct value. Bash here-strings (`<<<`) add a trailing newline that corrupts secrets.

中文简介正在准备中。当前显示原文说明。

divinevideo/divine-mobile2662026年10月19日 更新

Fix silent video/media processing failures caused by URL extraction code that filters on file extensions (.mp4, .webm, .webp). Use when: (1) Media moderation, transcoding, or analysis silently skips files from Blossom or content-addressed storage servers, (2) URL extraction from Nostr event tags (imeta, r tags) drops URLs without recognized extensions, (3) CDN fallback URLs append .mp4 but the actual server uses extensionless content-addressed paths like /{sha256}. Common in Nostr video events (kind 34236) where different clients use different URL formats.

中文简介正在准备中。当前显示原文说明。

divinevideo/divine-mobile2662026年10月19日 更新

查看 divinevideo 的全部技能

报告此技能的问题