Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Save Skill

ASecurity

Capture a reusable pattern from the current conversation into a local skill file under .claude/skills/local/ — LLM-draft + user-confirm, local-only, never synced. Use when the user says *save-skill, 'save this as a skill', '把这个存成 skill', or wants to keep a just-validated pattern.

3 stars
0 votes
0 copies
0 views
Added 10/6/2026
ai-agentsgogit

Works with

claude code

Security Analysis

A100/100

Scanned 10/6/2026

$npx -y skills add Sheldon-92/TAD --skill save-skill --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Save Skill?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Save Skill
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/sheldon-92-save-skill/badge)](https://www.skillsdirectory.com/skills/sheldon-92-save-skill)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: save-skill
description: "Capture a reusable pattern from the current conversation into a local skill file under .claude/skills/local/ — LLM-draft + user-confirm, local-only, never synced. Use when the user says *save-skill, 'save this as a skill', '把这个存成 skill', or wants to keep a just-validated pattern."
trigger: "*save-skill [optional name/topic hint]"
---

# *save-skill — Local Skill Capture

## Purpose

Bottom-up, in-the-moment capture: when a conversation just validated a small reusable
pattern, `*save-skill` turns it into a persistent, discoverable local skill file under
`.claude/skills/local/` — zero ceremony, no framework release. This COMPLEMENTS the Gate 4
Knowledge Assessment / distill loop (framework-level knowledge); it does not replace it.
Local skills are project-local tactical patterns, isolated from the TAD distribution chain.

平台绑定交互决策(cross-harness binding):本文件及其 references 中所有
AskUserQuestion 调用是「交互决策契约」而非具体工具——当前 harness 有该工具
(Claude Code)→ 直接调用;无该工具(Codex 等)→ 以编号纯文本列出全部选项
(1. … / 2. … / 3. …)并**停止等待用户输入**,用户以编号或自由文本作答;
禁止代答、禁止把选项折叠成默认值继续执行。SAFETY 门控的调用点(人工审批 /
归档确认 / 权限升级确认类)无论何种 harness 都必须获得真人作答后才能继续。
非交互执行模式(如 codex exec)→ 视为无人可答,按 blocked 停止并上报,
不得自选默认值;已按 YOLO/预授权模式运行且该决策点有书面预授权记录 →
按其协议处理,不适用本条 blocked 分支。

## Flow

Execute these steps in order. Steps 3 and 4 are BLOCKING — never skip them.

### Step 1 — Scan

Identify the ONE reusable pattern in the current conversation:

- What problem does it solve?
- What is the solution?
- When to apply it?
- When NOT to apply it?

If the user gave a hint (`*save-skill <hint>`), scope the scan to that hint.
If nothing capturable exists in the conversation, say so and STOP — do not invent a pattern.

### Step 2 — Draft

Fill the Local Skill Template (below). Naming rules:

- Name MUST be kebab-case matching `[a-z0-9-]+`.
- Derive the name from the pattern's TRIGGER (what future situation should recall it),
  not from the session topic.
- Target file: `.claude/skills/local/<name>.md`.
- `description` is one line containing the trigger keywords a future keyword-match will hit.

### Step 3 — Confirm [BLOCKING]

Show the FULL draft as a fenced markdown block.

🔒 MUST NOT write any file before the user confirms the draft

Offer the user (via AskUserQuestion or equivalent):

1. **Confirm write** — proceed to Step 4
2. **Edit draft** — user renames / trims / rewrites any part (taste and direction are
   human-domain judgments); apply the edits and loop back to showing the full draft
3. **Abort** — stop with zero file writes

Loop on edits until the user confirms or aborts.

### Step 4 — 🔒 OVERWRITE GUARD [BLOCKING]

If `.claude/skills/local/<name>.md` already exists, ask the user:

- **Overwrite** the existing file
- **Rename** the new skill (back to Step 3 with the new name)
- **Abort**

Never silently clobber an existing local skill.

### Step 5 — Write

1. `mkdir -p .claude/skills/local` (idempotent, on-demand — the directory is created at
   first write, never committed ahead of time).
2. Write `.claude/skills/local/<name>.md` per the confirmed draft.
3. If `.claude/skills/local/_index.md` does not exist, create it with this header:

   ```markdown
   # Local Skills Index

   > One line per local skill. Load path: read this index → match keywords → Read the file.
   > Format: `- [Title](<name>.md) — hook (max 120 chars)`

   ---
   ```

4. Append (or update in place, if the skill already has a line) the skill's index line:
   `- [Title](<name>.md) — hook (max 120 chars)`

File body and index line are written in the same Step 5 pass (file first, then index).
If an index line is ever found missing, re-running Step 5 repairs it idempotently.

### Step 6 — Report

Tell the user:

- The file path written (`.claude/skills/local/<name>.md`)
- The index line added to `_index.md`
- Reminder: the file is local-only — never synced, published, or distributed by TAD.

## Local Skill Template

```markdown
---
name: <kebab-case-name>
description: "<one line: what + trigger keywords>"
local: true
created: YYYY-MM-DD
source: save-skill
---

# <Title>

## When to use

## When NOT to use

## Steps

## Example

## Gotchas
```

## Using local skills

Load path for any agent in this project:

1. Read `.claude/skills/local/_index.md`
2. Keyword-match the task against the one-line hooks
3. Read the matched `.claude/skills/local/<name>.md` file(s)

## Constraints

- 🔒 MUST NOT write any file before the user confirms the draft
- 🔒 MUST only write under .claude/skills/local/ (plus `_index.md` in the same dir)
- 🔒 MUST NOT be auto-invoked — user-explicit command only (respects the blake skillify
  forbidden_implementations boundary: unattended skill materialization is forbidden; this
  flow only ever runs because the user explicitly asked for it)
- 🔒 local skills are never synced, published, or overwritten by TAD install
- Names MUST be kebab-case `[a-z0-9-]+`; every file MUST carry `local: true` frontmatter
- TAD-repo note: in the TAD framework repo itself `.claude/skills/local/` is gitignored
  (framework-repo-saved local skills never enter git → never enter the distribution clone →
  never get copied by tad.sh). In downstream projects, committing local skills is the
  project's own choice — tad.sh does not copy `.gitignore` and never touches
  `.claude/skills/local/` on install or update.

Attribution

Sheldon-92Sheldon-92
View sourceSee grades on GitHubMore from Sheldon-92 →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698621 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →