A community-driven registry for the Claude Code ecosystem. Not affiliated with Anthropic.
Are you the author? Sign in to claim
Claude Code skill — delegate coding tasks to Mistral Vibe, supervise via git diff & save 50-90% Claude token usage
Claude orchestrates. Vibe does the heavy lifting. You review the diff, save tokens, costs and avoid hitting limits!
Claude sees only ≈500–1500 tokens per run regardless of how many file reads Vibe performs internally — massive savings on exploratory and implementation tasks.
Note that Vibe works natively with Mistral models which are capable and significantly cheaper than Claude, but Vibe can also be configured to use any other provider/model instead. Eg you can use a deepseek model with vibe tooling.
Summary:
/vibe <instruction> in Claude Codevibe-delegate runs Mistral Vibe in a pseudo-TTYgit diff --statCost savings — Vibe's file reads and edits consume cheap delegate tokens (or whatever model you configure), not Claude tokens:
| Scenario | Claude Sonnet 4.6 | Mistral Medium 3.5 | DeepSeek V4 Flash |
|---|---|---|---|
| Simple 1-file tweak (800 tokens) | ≈$0.004 | ≈$0.002 | ≈$0.0001 |
| 6-read implementation task (4,800 tokens) | ≈$0.023 | ≈$0.012 | ≈$0.0008 |
| Complex multi-file refactor (12,000 tokens) | ≈$0.058 | ≈$0.029 | ≈$0.002 |
Costs based on official pricing (May 2026): Claude $3/$15 per M tokens, Mistral Medium 3.5 $1.50/$7.50, DeepSeek V4 Flash $0.14/$0.28. Assumes ≈85% input / 15% output, typical for coding tasks. Claude orchestration overhead: ≈500 tokens per run (negligible).
Le Chat Pro users: Mistral Vibe is included in the Le Chat Pro subscription (≈$18/mo). Mistral does not publicly document the exact usage limits, but community reports suggest ≈1–1.5B tokens/month are included. Within that allowance every delegation costs $0 in API fees — cheaper than any paid model.
Snapshot over 2,103 vibe delegations, 2026-05-12 → 2026-05-30 (19 days). Pulled from the shared run log via delegate-report (vibe scope).
Performance & savings
| Metric | Value |
|---|---|
| Delegations | 2,103 |
| Tokens delegated | 139.8M |
| Exit-success rate | 81% |
| Clean rate (no soft failure) | 67% |
| Avg run duration | 33s |
| Actually paid | ≈$0.67 DeepSeek (32M tokens) + Le Chat Pro sub (≈$18/mo) |
| Same workload on Claude Sonnet 4.6 | $456.94 |
| Effective saving | ≈ 24× cheaper than Claude |
By model
| Model | Runs | Exit-ok | Avg dur | Tokens | API-rate cost* | vs Claude |
|---|---|---|---|---|---|---|
| mistral-medium-3.5 | 1,718 | 79% | 31s | 104.2M | $137.09 | $206.24 |
| deepseek-flash | 269 | 93% | 41s | 32.1M | $35.43 | $67.12 |
| devstral-small | 68 | 63% | 19s | 2.6M | $0.26 | $7.81 |
| mistral (Le Chat) | 48 | 95% | 43s | 0.9M | $1.49 | $1.49 |
* Pay-as-you-go reference — mistral-medium ran on Le Chat Pro sub ($0 marginal); DeepSeek real cost was $0.67 (cache-aware), not $35.43.
deepseek-flash is the value pick (93% exit-ok at ≈$0.0025/run). devstral-small underperforms for edits — use it for read/explore only.
Error rate (real projects, 1,964 runs — synthetic test scaffolds excluded)
| Class | Rate | What it is |
|---|---|---|
| clean ok | 67.1% | completed and wrote files |
exit_error | 18.7% | engaged (≈7 tool calls) then exited non-zero, 95% wrote nothing |
wrote_nothing | 7.1% | tool calls but 0 files, exit 0 |
warn_only | 2.5% | non-fatal warnings, usually fine |
sr_fail | 2.4% | search_replace byte-match miss (accents, backticks) |
near_empty | 1.6% | <50 tokens out, nothing written |
syntax_error | 0.4% | wrote invalid code (caught by post-run gate) |
timeout | 0.3% | task too large / context saturated |
The 26% exit_error + wrote_nothing share one root cause: search_replace anchor not found byte-for-byte and the run abandons. The --require gate aborts these before they waste tokens.
Versus the other delegate (opencode)
| vibe | opencode | |
|---|---|---|
| Runs | 2,103 | 254 |
| Tokens | 139.8M | 8.5M |
| Exit-ok | 81% | 81% |
| API-rate cost (reference) | $174.28 | $0 (free tiers) |
| Models | mistral-medium, deepseek-flash (paid, capable) | free deepseek / mimo / nemotron tiers |
Same headline exit-rate, very different profile: opencode runs free tiers at zero cost but with high silent_exit and timeout rates. Use it for cheap exploration; vibe's paid models are the choice when the edit has to land. Both write to the same log — /vibe-report --all compares them.
vibe --version)script command available (GNU/Linux or BSD/macOS variant)timeout command available; on macOS install GNU coreutils for gtimeout (or ensure your chosen timeout fallback is set up)python3 and optionally node for syntax checksgit clone https://github.com/pcx-wave/vibe-skill.git && cd vibe-skill && mkdir -p ~/tools ~/.claude/skills/vibe ~/.claude/skills/vibeon ~/.claude/skills/vibeoff ~/.claude/skills/vibestatus ~/.claude/skills/vibe-model-pick ~/.claude/skills/vibe-model-clear ~/.claude/skills/vibe-report && ln -sf "$(pwd)/tools/vibe-delegate" ~/tools/vibe-delegate && ln -sf "$(pwd)/tools/delegate-report" ~/tools/delegate-report && chmod +x ~/tools/vibe-delegate ~/tools/delegate-report && ln -sf "$(pwd)/SKILL.md" ~/.claude/skills/vibe/SKILL.md && ln -sf "$(pwd)/VIBEON.md" ~/.claude/skills/vibeon/SKILL.md && ln -sf "$(pwd)/VIBEOFF.md" ~/.claude/skills/vibeoff/SKILL.md && ln -sf "$(pwd)/VIBESTATUS.md" ~/.claude/skills/vibestatus/SKILL.md && ln -sf "$(pwd)/VIBE-MODEL-PICK.md" ~/.claude/skills/vibe-model-pick/SKILL.md && ln -sf "$(pwd)/VIBE-MODEL-CLEAR.md" ~/.claude/skills/vibe-model-clear/SKILL.md && ln -sf "$(pwd)/VIBE-REPORT.md" ~/.claude/skills/vibe-report/SKILL.md
# 1. Clone this repo
git clone https://github.com/pcx-wave/vibe-skill.git
cd vibe-skill
# 2. Install the scripts (symlinks — stay in sync with git pull)
mkdir -p ~/tools
ln -sf "$(pwd)/tools/vibe-delegate" ~/tools/vibe-delegate
ln -sf "$(pwd)/tools/delegate-report" ~/tools/delegate-report
chmod +x ~/tools/vibe-delegate ~/tools/delegate-report
# 3. Install the skills for Claude Code
mkdir -p ~/.claude/skills/vibe ~/.claude/skills/vibeon ~/.claude/skills/vibeoff ~/.claude/skills/vibestatus \
~/.claude/skills/vibe-model-pick ~/.claude/skills/vibe-model-clear ~/.claude/skills/vibe-report
ln -sf "$(pwd)/SKILL.md" ~/.claude/skills/vibe/SKILL.md
ln -sf "$(pwd)/VIBEON.md" ~/.claude/skills/vibeon/SKILL.md
ln -sf "$(pwd)/VIBEOFF.md" ~/.claude/skills/vibeoff/SKILL.md
ln -sf "$(pwd)/VIBESTATUS.md" ~/.claude/skills/vibestatus/SKILL.md
ln -sf "$(pwd)/VIBE-MODEL-PICK.md" ~/.claude/skills/vibe-model-pick/SKILL.md
ln -sf "$(pwd)/VIBE-MODEL-CLEAR.md" ~/.claude/skills/vibe-model-clear/SKILL.md
ln -sf "$(pwd)/VIBE-REPORT.md" ~/.claude/skills/vibe-report/SKILL.md
# 4. (Optional) Enable auto-mode — Claude delegates all code tasks automatically
# without requiring /vibe each time. Toggle with /vibeon and /vibeoff.
grep -q "vibe auto-mode" ~/.claude/CLAUDE.md 2>/dev/null || cat >> ~/.claude/CLAUDE.md << 'EOF'
# vibe auto-mode
At the start of every user request that involves writing, editing, or fixing code:
1. Run `test -f ~/.local/share/vibe-auto.flag` (silent, no output to user).
2. If the flag exists → automatically invoke the `vibe` skill exactly as if the user had typed `/vibe <their full instruction>`. Do NOT ask first, do NOT explain — just delegate.
3. If the flag is absent → proceed normally.
The flag is toggled by `/vibeon` and `/vibeoff`.
EOF
Verify with ~/tools/vibe-delegate /tmp "Say hello in one sentence." 3
Because both installs use symlinks, a git pull is all you need:
cd ~/vibe-skill && git pull
~/tools/vibe-delegate and ~/.claude/skills/vibe/SKILL.md are automatically up to date — no re-copy needed.
Migrating from a previous
cp-based install? Replace the copies with symlinks:hljs language-bashcd ~/vibe-skill ln -sf "$(pwd)/tools/vibe-delegate" ~/tools/vibe-delegate ln -sf "$(pwd)/SKILL.md" ~/.claude/skills/vibe/SKILL.md
In a Claude Code session:
/vibe add a dark mode toggle to the settings page
/vibe the login form is not validating the email field — fix it
/vibe add pagination to the GET /posts route, 20 items per page
Claude decomposes the task, writes the Vibe prompt, supervises execution, and reports the diff.
By default, Vibe uses whatever active_model is set in ~/.vibe/config.toml. You can override it per-session without touching that file:
/vibe-model-pick — interactive menu built from your config.toml models
/vibe-model-pick devstral-small — switch directly by alias
/vibe-model-clear — remove the override, return to config default
/vibestatus — shows both auto-mode state and active model override
The override is stored in ~/.local/share/vibe-model.flag and is picked up by vibe-delegate on every run. It persists across sessions until you clear it.
For frictionless delegation, enable auto-mode once in your Claude Code session:
/vibeon — every code request is automatically delegated to Vibe, no /vibe prefix needed
/vibeoff — return to normal Claude behaviour
/vibestatus — auto-mode state + active model override
With vibeon active, just talk to Claude normally:
add pagination to the /posts route
fix the broken email validation
refactor the auth middleware into its own module
Claude intercepts code requests and applies a pre-filter before delegating:
| Task | Action |
|---|---|
| 1 file, ≤10 lines, exact location known | Claude edits directly — no delegation overhead |
| Everything else | Delegated to Vibe |
Pure questions and conversations always go directly to Claude.
Sample output from a real run:
=== VIBE START ===
Workdir : /path/to/project
Agent : default
Model : (config default)
Turns : 10
Timeout : 180s
Prompt : Stack: Python/Flask. File: app.py ...
===================
[read] app.py
[tool] file: app.py
[tool] search_replace [OK] ...
[vibe] Done. Converted date to datetime.date in fetch_data().
Tool calls: 5 | warns: 0 | sr_fails: 0
Model : deepseek-flash
Delegate tokens (run): 4,800 (last turn: 4,600+200) | cost ~$0.0007
Claude Sonnet 4.6 eq: same tokens would cost ~$0.0168 (ratio x24.0)
=== VIBE DONE (exit: 0) ===
=== SYNTAX OK (1 file(s) checked) ===
=== UNCOMMITTED CHANGES ===
app.py | 4 ++--
[log] → ~/.local/share/delegate-runs.jsonl (4800 tokens, exit 0, 34.2s, saved ~$0.0161 vs Claude)
Claude Code
└─ /vibe <instruction>
└─ SKILL.md logic
└─ ~/tools/vibe-delegate <workdir> <prompt> [turns] [agent] [timeout]
├─ writes prompt to temp file (avoids shell injection with UTF-8/emoji)
├─ generates a temp shell script for the vibe command
├─ runs: python3 -c 'pty.spawn(<vibe-script>)'
│ └─ allocates pseudo-TTY (required — vibe hangs without one)
├─ pipes JSON streaming output through Python parser
│ └─ prints [read] / [write] / [WARN] / [vibe] lines
├─ reads real token counts from Mistral session log
├─ runs syntax checks on modified .py and .js files
├─ prints git diff --stat
└─ appends JSON entry to ~/.local/share/delegate-runs.jsonl
Cost figures are estimates. See
SKILL-reference.mdfor methodology.
examples/good-prompts.md — prompt patterns that reliably workexamples/anti-patterns.md — what fails and why, with fixesA parallel delegate using Gemini CLI is available at pcx-wave/gemini-skill. Same orchestration pattern, same run log format — different model and trade-offs.
Every run is logged to ~/.local/share/delegate-runs.jsonl with tokens, cost, model, and failure details. Query it with ~/tools/delegate-report [--since N] [--project NAME] [--fails] or from Claude Code: /vibe-report [args].
The log is shared with sister delegates (e.g. gemini-skill), so the report defaults to vibe runs only. Use --all for the cross-delegate comparison, or --delegate NAME to scope to another tool.
See docs/feedback-claude-sonnet.md for original feedback from Claude after hours of practice that drove the iterations on vibe-delegate — real bugs hit, root causes, and the fixes applied.
MIT
ML engineering — model training, deployment, MLOps, monitoring
DevOps practices — CI/CD, containers, monitoring, infrastructure automation
Professional skills marketplace with production-ready skills for enhanced development
Self-learning system that captures corrections and syncs them to CLAUDE.md and AGENTS.md
