Borrowing it
Nothing to install: this file belongs to ashishpatill/tell-proof. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ashishpatill/tell-proof/master/.cursor/skills/tell-recursive-improve/SKILL.mdgit clone --depth 1 https://github.com/ashishpatill/tell-proofWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/ashishpatill/tell-proof/tell-recursive-improve)<a href="https://agentmods.dev/skills/ashishpatill/tell-proof/tell-recursive-improve"><img src="https://agentmods.dev/badge/skills/ashishpatill/tell-proof/tell-recursive-improve/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/ashishpatill/tell-proof/tell-recursive-improve"><img src="https://agentmods.dev/badge/skills/ashishpatill/tell-proof/tell-recursive-improve.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00081 | $0.00919 |
| Opus 5 | $0.00041 | $0.00460 |
| Sonnet 5 | $0.00016 | $0.00184 |
| Haiku 4.5 | $0.00008 | $0.00092 |
Grade A, and why
tell-recursive-improve 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 6d 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.
How it starts
The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tell recursive improve (champion → challenger → learnings)
Compounding loop. Each session must leave one durable lesson and one measurable win. Do not brainlessly re-shoot the same nav strip and call it proof.
Harness
| Gate | Fail if |
|---|---|
| Eye — showcase | Preview shows sticky nav + truncated type; empty right gutter; ghosted labels |
| Eye — template | Best craft moment (figure / inverse / claim) not visible in posted shots |
| Eye — measured contrast | pnpm eye:shell red, or chrome text looks "fine" only in a vision caption (vision ≠ WCAG) |
| Critique | Matrix or target brief score regresses |
| Basics | assertBasics red |
Pass bar for showcase shots: the frame must include a primary craft beat (hero claim with figure start, or product plate, or inverse proof) — not chrome alone.
Pass bar for shell/chrome: run pnpm eye:shell (Playwright samples computed colors). Do not ship sidebar/rail changes on agent vision alone - captions miss ghost contrast (chrome:rail-ghost-contrast).
Loop (one cycle)
- Load learnings — read
research/LEARNINGS.mdbefore editing. - Name the failure — one sentence (e.g. "index thumbs are 88px tall at scale 0.2 → only nav visible").
- Champion — current committed behavior.
- Challenger — smallest change that targets that failure (one surface: preview crop, gallery UI, or one siteKind).
- Eval — Playwright eye on
/showcase+ offering;pnpm research:critiqueif engine touched; typecheck. - Promote or revert — keep only if eye + critique clear.
- Write the lesson — append to
research/LEARNINGS.mdwith pattern key + anti-pattern. - Screenshot contract — post gallery + at least one mid-page craft beat (or cinema reel frame), never nav-only.
Pattern keys (examples)
showcase:preview-gutter— fixed iframe scale left dead stripshowcase:nav-only-thumb— short crop at y=0template:label-under-claim— absolute overfigure under typetemplate:sitekind-css-dead— siteKind rules trapped in wrong lean branchtemplate:skill-metadata-only— skills listed in routedSkills/hints but not hard-gated; requireassertSkillWiring+ RESEARCH_GATE.md every agency 2-buildchrome:rail-ghost-contrast— paper sidebar using dark-theme light text; requirepnpm eye:shell
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.
- 6d ago First seen · 70 lines · 81 tokens per session scan A df66f3299ffb
tell-recursive-improve is a skill published in the GitHub repository ashishpatill/tell-proof (1 stars, last pushed 15d ago), licensed MIT. It adds 81 tokens to every session and 919 once invoked, about $0.0004 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 2026-09-03.
Other skills, from other repositories
ijfw-ui-spec
Use when the user says: 'ui spec', 'design contract', 'ui audit setup', 'lock the design', 'visual contract', 'ui review setup', or '/ijfw-ui-spec'. Produces UI-SPEC.md as the visual design contract before any frontend or visual-artifact build, and dispatches ijfw-ui-auditor as the final 6-pillar gate.
ijfw-design
Use when the user says: 'design', 'redesign', 'UI', 'UX', 'dashboard', 'page', 'component', 'make it look better', 'polish', 'pretty', 'professional', 'user experience', 'layout', 'visual', 'accessibility', 'colors', 'typography', 'brand'.
better-design
Build, improve, and review production interfaces with the Better Design MCP. Use for frontend design, UI, UX, interactions, flows, forms, navigation, components, styling, design systems, accessibility, responsive layouts, motion, React Native, Expo, Shopify, Three.js, and product video work, especially when creating…
Add Claude design-system and accessibility workflows with UX/UI Agent Skills
Use UX/UI Agent Skills when Claude should generate tokens, component specs, accessibility audits, and framework-specific UI code from a repeatable design workflow.
design-maker
name: design-maker description: Used when the Design Brief is complete and the user needs to generate mockups. Reads Product-Spec.md and Design-Brief.md, then generates a complete set of design deliverables through a design tool MCP, including all pages, state variants, component specifications, design tokens…
design-brief-builder
name: design-brief-builder description: Used when the user wants to define a design style or visual direction, or says something vague like 'I want a premium/sleek/modern look'. Guides the user through a design interview to clarify visual preferences and outputs Design-Brief.md. After Brief is saved, MUST run Next…