Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/saski/arnesto/collaborative-designnpx skills add saski/arnesto --skill collaborative-designgit clone --depth 1 https://github.com/saski/arnestoWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00035 | $0.00608 |
| Opus 5 | $0.00017 | $0.00304 |
| Sonnet 5 | $0.00007 | $0.00122 |
| Haiku 4.5 | $0.00003 | $0.00061 |
Grade A, and why
collaborative-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
This is a copy
100% identical to collaborative-design — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
STARTER_CHARACTER = 🎨
Purpose
Stay in design mode. Resist jumping to implementation. Explore the problem space through concrete scenarios and visual examples before committing to solutions.
Core Principle: Show, Don't Tell
Instead of describing what happens in prose, show it visually. Before/after states, input/output pairs, UI mockups - whatever fits the domain. Visual beats prose.
Bad: "The system strips Claude co-authors and keeps human ones" Good: Show the actual input and expected output side by side
Process
Problem → Research → Timeline → Scenarios → Decisions → Validation
↑__________________________________________________|
(iterate freely)
1. Clarify the Problem
- What are we building? Why?
- What does success look like?
- What constraints exist?
2. Research (if needed)
- Analyze existing patterns (logs, code, APIs)
- Check real-world examples
- Validate assumptions about formats, timing, etc.
3. Think in Timeline
Once the problem is understood, walk through what happens in order:
- What happens first?
- Then what?
- What triggers the next step?
This uncovers unknowns. Each step becomes a scenario to explore.
4. Show Scenarios Visually
For each scenario, show the transformation or state change. Format depends on domain:
- Config/data: show before and after
- UI: show screen states and transitions
- Pipelines: show input → output
- Workflows: show steps with arrows
Use domain language. Stay high-level. Show complete examples, don't truncate.
5. Surface Options, Then Decide
Present several options with tradeoffs. Don't decide alone:
- "Option A does X, Option B does Y. Which direction?"
- Wait for input before proceeding
6. Validate Before Building
- POC for risky assumptions (API timing, format parsing)
- Visual test cases (input → expected output)
- Document findings
7. Document Decisions
Track what was decided and why. Update docs as design evolves.
Anti-patterns
- Jumping to code before exploring the design space
- Describing scenarios in prose instead of showing them visually
- Showing one solution instead of options
- Asking multiple questions at once (show the whole list, then ask each one at a time)
- Making assumptions without checking (real data, real APIs)
- Skipping visuals for "obvious" cases
- Truncating examples (show complete data)
- Deciding without discussing tradeoffs
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 87 lines · 35 tokens per session scan A 0fa5c84d8989
collaborative-design is a skill published in the GitHub repository saski/arnesto (5 stars, last pushed 7d ago), licensed Unlicense. It adds 35 tokens to every session and 608 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to collaborative-design, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
pixir-delegate
Use Pixir as a headless subagent runtime from Claude Code or any harness with skill ! preprocessing (Codex roots and other no-hydration hosts use pixir-delegate-codex instead) — one-shot workers (pixir --json), parallel fan-out to N children (pixir delegate --spec), resumable steering (pixir resume), evidence…
pixir-delegate-codex
Use when a Codex CLI/Desktop root should fan out subagents, delegate to Pixir workers, run parallel workers, or manage a resident delegation daemon via Pixir Delegate; covers Codex preflight, AGENTS.md, approvals/sandbox, dry-run, daemon start/status/attach/cancel, closure evidence, and audited single-run execution…
pixir-diagnostics
Diagnose Pixir and T3 Code Pixir incidents from local canonical evidence. Use when a Pixir run, ACP/T3 thread, subagent/workflow, provider replay, or daily-driver dogfood session appears stuck, inconsistent, missing tool output, or hard to classify.
readonly-review
Run a no-network read-only review practice with two explorer steps and one synthesis step.
pixir-delegate-native
Delegate work to subagents from INSIDE a Pixir session using the native Subagent tools (spawnagent, waitagent, closeagent, listagents, sendinput) instead of shelling out to the pixir CLI. Use when you are a Pixir session that needs to fan out parallel workers, steer a child, or run skill-backed workflow templates …
repo-contracts-and-boundaries
Use when turning architecture, layering, ownership, dependency direction, schemas, structural metrics, quality thresholds, baselines, allowlists, or generated quality snapshots into repository checks.