scratchpad

A persistent note that records attempts, failures, results, and possible next steps during an iterative task.

In plain words
What is it for?
Use it for debugging, experiments, and other tasks that require trying several approaches over time.
Why use it?
It prevents an agent from repeating approaches that already failed, including after the task is paused or resumed.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/5uck1ess/devkit/scratchpad
Any agent
npx skills add 5uck1ess/devkit --skill scratchpad
Clone the repo
git clone --depth 1 https://github.com/5uck1ess/devkit

Made for: Claude Code, Codex.

Per session 27 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 429 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00027 $0.00429
Opus 5 $0.00014 $0.00215
Sonnet 5 $0.00005 $0.00086
Haiku 4.5 $0.00003 $0.00043

Measured yesterday against content hash ca65e441359d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

scratchpad 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 yesterday.

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.

skills/scratchpad/SKILL.md · 57 lines

What it actually says

Scratchpad Protocol

Use scratchpads in any iterative loop to prevent repeating failed approaches.

Location

.devkit/scratchpads/current.md — active scratchpad for the current task.

Create .devkit/scratchpads/ if it doesn't exist. Only one active scratchpad at a time.

Before Each Iteration

Read .devkit/scratchpads/current.md if it exists. Check:

  • What approaches were already tried?
  • What failed and why?
  • What was suggested to try next?

Do not repeat a failed approach. If you're about to try something already listed as failed, stop and pick a different strategy.

After Each Iteration

Append to .devkit/scratchpads/current.md:

## Iteration {N} — {timestamp}

**Approach:** What you tried (one sentence)
**Result:** pass | fail
**Details:** What happened — error message, unexpected behavior, or success details
**Next:** What to try next if this failed, or "N/A" if it passed

On Completion

When the task succeeds or the workflow ends, delete .devkit/scratchpads/current.md. Don't leave stale scratchpads — they'll confuse the next task.

Integration with stuck detection

If .devkit/scratchpads/current.md shows 3+ failed iterations:

  1. Stop iterating
  2. Review all failed approaches in the scratchpad
  3. The pattern of failures often reveals the real problem
  4. Escalate to the user with the scratchpad content as evidence

Rules

  • One scratchpad per active task — don't create per-agent scratchpads
  • Keep entries concise — the scratchpad is read every iteration
  • Record failures honestly — "it didn't work" is useless; "returned 404 because endpoint expects POST not GET" is useful
  • Clean up when done — stale scratchpads are worse than no scratchpads
Changes

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.

  1. yesterday First seen · 57 lines · 27 tokens per session scan A ca65e441359d

Subscribe to this mod's changes

scratchpad is a skill published in the GitHub repository 5uck1ess/devkit (5 stars, last pushed 13d ago), licensed MIT. It adds 27 tokens to every session and 429 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories