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 skills add isvlasov/rageatc-oss --skill debloatinggit clone --depth 1 https://github.com/isvlasov/rageatc-ossWrote 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/isvlasov/rageatc-oss/debloating)<a href="https://agentmods.dev/skills/isvlasov/rageatc-oss/debloating"><img src="https://agentmods.dev/badge/skills/isvlasov/rageatc-oss/debloating.svg" alt="Measured on agentmods" 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 | $0.00078 | $0.01130 |
| Opus 5 | $0.00039 | $0.00565 |
| Sonnet 5 | $0.00016 | $0.00226 |
| Haiku 4.5 | $0.00008 | $0.00113 |
Grade A, and why
debloating 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 8d 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 — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debloating
What counts as bloat
Bloat is text that changes nothing for its reader. The reader is usually the agent, but not always — Step 0 confirms who. Text fails it when: the reader already knows it, it is filler or ceremony, it repeats something already said.
This is judged against the artefact's intent. Without an agreed intent, signal and noise are indistinguishable.
Scope
- The unit is one file. Read related assets (another skill, a rule, the agent that preloads this skill) only when a specific cut needs that context — for example, to confirm a passage is duplicated elsewhere or already known to the reader. Most files need none of this.
- Tuned for skills; works on any prose artefact.
- Frontmatter descriptions are in scope — they are often bloated. Debloat them too, but preserve the trigger keywords and discoverability that let the skill be found; do not cut past the point where routing still works.
- Flag, do not fix, anything that is not a cut: contradictions, overlap with another skill, aged assumptions, "should this exist at all". Raise it and let the user decide.
Workflow
Step 0 — Agree the intent (always first)
Read the whole artefact. State back, and confirm with the user, three things: why it exists (its intent), who the reader is (usually the agent, sometimes a human), and what matters to that reader. These are the measuring stick for every judgement that follows — cut nothing before they are agreed.
Step 1 — Classify every passage
Sort candidate text by what removing it costs:
- Tier 1 — lossless removal. Text that fails the test above: no signal. Filler, ceremony, internal repetition, explaining what the reader already knows.
- Tier 2 — lossless compression. The instruction stays; the ornament goes. Same load, shorter and plainer words.
- Tier 3 — lossy reduction. Real but marginal signal, removed because the text-to-value ratio is poor — a small loss for a large saving.
- Offload. Detail that is genuinely needed but only sometimes (edge cases, long schemas, example catalogues) moves to
references/with a one-line pointer. Rule: what is needed on every invocation stays in the body; what is needed only conditionally is relocated, not deleted.
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.
- 8d ago First seen · 63 lines · 78 tokens per session scan A 03a4824d9541
debloating is a skill published in the GitHub repository isvlasov/rageatc-oss (9 stars, last pushed 1mo ago), licensed MIT. It adds 78 tokens to every session and 1,130 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 2026-08-31.
Other skills, from other repositories
design-method
Use when starting any design task — resolves register, sets the 3 dials, locks Gate 0, and routes to the right move and target skill.
design-webapp
Use when building or extending a dashboard, SaaS screen, or any authenticated app surface — use design-web instead for a marketing site.
design-web
Use when building or revising a marketing/landing/campaign page, register brand locked at Gate 0 — use design-webapp for app screens instead.
laravel-architecture
Use when structuring a Laravel project, creating services/repositories/actions, implementing dependency injection, or organizing code layers.
laravel-billing
Use when implementing subscriptions, invoices, payment methods, webhooks, or billing portals with Laravel Cashier (Stripe or Paddle).
laravel-permission
Use when implementing RBAC, role-based access control, or user authorization with Spatie Laravel Permission.