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 IgorWarzocha/howaboua-pi-stuff --skill prompt-cachinggit clone --depth 1 https://github.com/IgorWarzocha/howaboua-pi-stuffWrote 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/igorwarzocha/howaboua-pi-stuff/prompt-caching)<a href="https://agentmods.dev/skills/igorwarzocha/howaboua-pi-stuff/prompt-caching"><img src="https://agentmods.dev/badge/skills/igorwarzocha/howaboua-pi-stuff/prompt-caching/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/igorwarzocha/howaboua-pi-stuff/prompt-caching"><img src="https://agentmods.dev/badge/skills/igorwarzocha/howaboua-pi-stuff/prompt-caching.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00022 | $0.01560 |
| Opus 5 | $0.00011 | $0.00780 |
| Sonnet 5 | $0.00004 | $0.00312 |
| Haiku 4.5 | $0.00002 | $0.00156 |
Grade A, and why
prompt-caching 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 10d 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 — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Define success
Successful prompt caching reuses naturally stable model context without compromising what the model should receive. Cold requests should occur at understandable lifecycle boundaries. Later requests should read the expected stable prefix, process the changed tail normally, and repay writes through real cost or latency savings.
Do not optimize for the highest read percentage. The useful boundary is the longest prefix that should honestly remain unchanged.
Keep these mechanisms separate:
- provider prompt caching reuses model prefill for an exact rendered prefix
- server continuation carries conversation state between requests
- transport prewarm prepares an expected request or connection
- cache prediction estimates reuse but does not inspect provider cache state
A key, stable source object, fast response, prediction, or successful continuation does not prove a provider hit. Use provider-reported reads and writes as final evidence.
Trace lifecycle and every model call
Cache correctness has two axes: payload identity and snapshot authority. Discover the harness's actual event order, including session start or resume, repeated turn preparation, provider calls, tool continuations, retries, lane changes, navigation, and compaction where present.
- Map every route that can invoke the model. Ordinary input, extension commands, model-visible messages, synthetic user messages, queued follow-ups, steering, retries, and continuations may run different hooks. Do not infer hook coverage from message role or visibility.
- Separate adding context from starting work. If work requires fresh prompt or tool state, use a route that performs authoritative preparation.
- Treat an idle check as an observation, not a reservation. Revalidate at invocation or use an atomic route when freshness depends on a new prepared run.
- Establish each request, continuation, replay, prewarm, or compaction snapshot's session and lane, capture event, completed hooks, represented provider request, consumer, and invalidators. The authoritative baseline is normally the latest completed final provider request in the current lane. Session-start state is not timeless.
- Guard asynchronous state so late work from an old lane cannot overwrite newer state.
- Record every provider call chronologically. An agent turn may contain several. Capture its trigger, lane, continuation state, uncached input, cache reads, cache writes, snapshot provenance, expected stable prefix, and first divergence.
- Explain the sequence before aggregating it. A series of misses followed by one large hit is broken until explained. Long-term ratios hide when, where, and for whom caching failed.
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.
- 10d ago First seen · 97 lines · 22 tokens per session scan A d88fc770dfa9
prompt-caching is a skill published in the GitHub repository IgorWarzocha/howaboua-pi-stuff (359 stars, last pushed yesterday), licensed MIT. It adds 22 tokens to every session and 1,560 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 2026-08-30.
Other skills, from other repositories
llm-app-patterns
Production-ready patterns for building LLM applications. Covers RAG pipelines, agent architectures, prompt IDEs, and LLMOps monitoring. Use when designing AI applications, implementing RAG, building agents, or setting up LLM observability.
prompt-optimization
Improve a prompt on the evaluations workbench through a measured loop. Score the baseline first, then duplicate the target column, form a hypothesis from failing rows, edit the copy's prompt draft, run, compare pass rate and cost, and repeat until the numbers hold. Use when the user asks to optimize or improve a…
enhance-prompt
Transforms vague UI ideas into polished, Stitch-optimized prompts. Enhances specificity, adds UI/UX keywords, injects design system context, and structures output for better generation results.
prompt-engineer
Writes, refactors, and evaluates prompts for LLMs — generating optimized prompt templates, structured output schemas, evaluation rubrics, and test suites. Use when designing prompts for new LLM applications, refactoring existing prompts for better accuracy or token efficiency, implementing chain-of-thought or few-shot…
seedance-vocab-en
This skill should be used when an English Seedance 2.0 prompt needs clearer production wording, less generic prose, or precise vocabulary for camera, lighting, motion, VFX, audio, and constraints. Route blocked prompts through seedance-filter for context and boundary review.
ideogram4
Prompting patterns for Ideogram 4 text-to-image — best-in-class in-image text rendering and exact color/layout control via structured JSON captions. Use when generating images that need legible on-image text (title cards, thumbnails, logos, signage, CTAs), precise brand colors, or controlled spatial layout. Triggers…