dead-ends — Claude Code skill: what it does, how to install
dead-ends is a skill that shows approaches a repository tried and abandoned, with reasons, and flags when current work heads toward a recorded dead end.
MIT ★ 586 updated 2026-10-26
/plugin marketplace add karanb192/claude-code-hooks
/plugin install dead-end-registry@claude-code-hooksWhat dead-ends does
The skill displays repository approaches that were tried and later reverted or ruled out. Entries are ordered newest first and include when the attempt happened, why it was abandoned, and its estimated token cost. This is a record of decisions in that repository, not general advice.
The dead-end-registry hook builds the list by mining work transcripts. The skill briefly summarizes one to three relevant entries and flags a past detour when it relates to the current task; it does not rerun commands. To install through the marketplace, run both commands in order. If the registry has no entries, the skill says the list is empty.
What it helps with
- find approaches already abandoned in a repository
- understand why an earlier approach was dropped
- check the estimated token cost of a detour
- spot when current work repeats a recorded dead end
Example prompts for the agent
- Show me the approaches this repository tried and later abandoned.
- Check the dead-end records and tell me if this task repeats one.
- Summarize recent failed approaches and why they were dropped.
How to install dead-ends
1Plugin marketplace
/plugin marketplace add karanb192/claude-code-hooks/plugin install dead-end-registry@claude-code-hooksThis installs the whole dead-end-registry plugin with all its skills
2Skills CLI (Vercel)
npx skills add karanb192/claude-code-hooks --skill dead-ends -a claude-code3Manually: put the files in place
~/.claude/skills/dead-ends/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/dead-ends/
1Skills CLI (Vercel)
npx skills add karanb192/claude-code-hooks --skill dead-ends -a codex2Manually: put the files in place
.agents/skills/dead-ends/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add karanb192/claude-code-hooks --skill dead-ends -a cursor2Manually: put the files in place
.cursor/skills/dead-ends/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install karanb192/claude-code-hooks plugins/dead-end-registry/skills/dead-ends2Manually: put the files in place
.github/skills/dead-ends/SKILL.md1Skills CLI (Vercel)
npx skills add karanb192/claude-code-hooks --skill dead-ends -a gemini-cli2Platform CLI
gemini skills install https://github.com/karanb192/claude-code-hooks --path plugins/dead-end-registry/skills/dead-endsManually: put the files in place
.devin/skills/dead-ends/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about dead-ends
How do I install dead-ends in Claude Code?
Run these commands in order: /plugin marketplace add karanb192/claude-code-hooks, then /plugin install dead-end-registry@claude-code-hooks.
What does the dead-ends skill do?
It shows repository approaches that were reverted or ruled out, with reasons. Entries also include the attempt date and estimated token cost.
Where do the dead-end records come from?
The dead-end-registry hook mines work transcripts to produce them. The skill reads the registry and does not rerun commands to rebuild it.
Source text
SKILL.md · MIT
---
name: dead-ends
description: List this repo's recorded dead ends: approaches already tried and abandoned, with reasons and cost
disable-model-invocation: true
---
# Dead ends
!`node "${CLAUDE_SKILL_DIR}/../../dead-end-registry.js" --render 2>/dev/null || node "${CLAUDE_PLUGIN_ROOT}/dead-end-registry.js" --render`
The list above is this repository's recorded dead ends: approaches that were tried
and then reverted or ruled out, newest first, each with the date it was tried, why it
was walked back, and the estimated token cost of the detour. It is produced by the
`dead-end-registry` hook, which mines your transcripts as you work.
Briefly summarize the 1-3 most relevant entries and, if any bear on what the user is
currently doing, gently flag that they have been down this road before. If the list is
empty, just say so. Keep it short; do not re-run any commands.
© Karan Bansal · MIT · karanb192/claude-code-hooks