Borrowing it
Nothing to install: this file belongs to Lengros/bennu. 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/Lengros/bennu/main/.claude/skills/speech/SKILL.mdgit clone --depth 1 https://github.com/Lengros/bennuWrote 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/lengros/bennu/speech)<a href="https://agentmods.dev/skills/lengros/bennu/speech"><img src="https://agentmods.dev/badge/skills/lengros/bennu/speech/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/lengros/bennu/speech"><img src="https://agentmods.dev/badge/skills/lengros/bennu/speech.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 | $0.00195 | $0.01966 |
| Opus 5.5 | $0.00078 | $0.00786 |
| Sonnet 5.5 | $0.00039 | $0.00393 |
| Haiku 4.5 | $0.00019 | $0.00197 |
Grade A, and why
speech 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 2mo 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 — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/speech — dictation → delivery-ready spoken script
Takes raw, dictated, out-of-order ideas (often voice-to-text, often in one language for delivery in another) and returns a script the user can read aloud. It does the structuring, translation, and verbal polish — it does not invent content the speaker didn't give.
The engine, repeated at every stage: this text will be spoken, not read. Every choice serves the ear and the speaker's authentic voice — never the page. Preserve the speaker's own honesty (tentative claims stay tentative) and deliberate humor/attribution; written-doc register and pitch-deck clichés are the failure modes.
Argument (optional): a topic/title, or pre-existing notes. If absent, the user is about to dictate — set up to capture (§2) and wait.
This is artifact-shaped work: delegate the writer/editor passes to fresh subagents, each
embodying a purpose-built persona defined inline below (a narrative_writer, then a
narrative_editor) — inline production reads flat. The orchestrator runs intake + capture +
delivery; the personas produce the prose.
How to run
1. Intake — settle the load-bearing parameters BEFORE capture
The first three shape the whole script and must not be guessed — ask (AskUserQuestion is ideal) unless already stated. The fourth (register) is derived, not asked.
- Language of the final script (dictation language may differ → translation needed).
- Audience — who's in the room (internal team / clients / investors / conference / public). Drives register and what can be assumed vs explained.
- Length in minutes → convert to a word budget at ~130–150 words/min (e.g. 5 min ≈ 650–750 words). The budget is a ceiling — cut to fit.
- Register — live demo-walkthrough (first person, "let me show you") vs polished talk vs internal update. Infer it from Audience + Length rather than asking a separate question: short internal talk → loose first-person; investor/keynote/press → tighter and more composed. Only ask if genuinely ambiguous. Never over-structure a casual talk.
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.
- 2mo ago First seen · 126 lines · 0 tokens per session scan A 4e729ea3a7ff
speech is a skill published in the GitHub repository Lengros/bennu (2 stars, last pushed 3mo ago), licensed MIT. It adds 195 tokens to every session and 1,966 once invoked, about $0.0008 per session on Opus 5.5. 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
dcf-model
Build discounted cash flow valuation workbooks in Excel.
mle-workflow
Production machine-learning engineering workflow for data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback. Use when building, reviewing, or hardening ML systems beyond one-off notebooks.
autonomous-loops
Patterns and architectures for autonomous Claude Code loops — from simple sequential pipelines to RFC-driven multi-agent DAG systems.
grounded-citations
Ground answers and documents in cited, verifiable sources.
hermes-agent-skill-authoring
Author in-repo SKILL.md files: frontmatter and structure.
systematic-debugging
4-phase root cause debugging: understand bugs before fixing.