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 agentmods add skills/stainless-code/codemap/docs-voicenpx skills add stainless-code/codemap --skill docs-voicegit clone --depth 1 https://github.com/stainless-code/codemapWhat 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 | $0.00065 | $0.01726 |
| Opus 5 | $0.00032 | $0.00863 |
| Sonnet 5 | $0.00013 | $0.00345 |
| Haiku 4.5 | $0.00006 | $0.00173 |
Grade A, and why
docs-voice 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.
How it starts
The opening of the file, as written. The whole thing — 122 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Docs voice — Codemap docs (apps/docs, built with Blume)
Keep landing, guides, concepts, recipes, reference, and generated API reading like one voice.
Voice in one line
Senior-dev to senior-dev: concrete, SQL-literate, dry, honest about scope. No hype. The differentiators are radical honesty (say what's planned / diverges / was rejected) and the "when not to use it" anti-pitch — keep both.
Do
- Lead with the pain, then the mechanism. "Agents burn tokens scanning files
to find one symbol…" → then "
codemap query --jsonreturns the row in one SQL round-trip." - Concrete before abstract. Name
codemap query/--recipe/createCodemap/defineConfigbefore the coinage "predicate-as-API" or "structural index." - CLI-first AND programmatic. Show the CLI shape first; pair with the library
shape (
createCodemap,defineConfig, adapters) when the page is API-shaped. The published package entry is the SSOT for generated/reference/api(TypeDoc). - SQL is the API. When a page answers a structural question, show the SQL or
the
--recipe— not a paragraph of prose pretending to be the answer. Recipes are named SQL patterns; cite the recipe id. - Section headers by page type. Marketing = period-terminated benefit sentence ("Query your codebase."); guides = action verb ("Find a symbol with one query"); reference = precise noun; concepts = model noun + consequence.
- Card titles = the outcome; card bodies = the API / SQL.
- One idea per sentence in leads. Short claim first, then expand.
- State experimental / pre-1.0 status once per surface, one wording (see Canonical patterns).
- Sidebar icons: all-or-none per sibling list. Blume does not reserve an
icon column — sparse
sidebar.iconjaggeds labels. Guides, Concepts, Recipes, Reference leaves: none unless every peer has a natural glyph. Sectionmeta.tsicons and tab icons stay. - Peer framing is structural, not brand-vs-brand. When comparing, contrast
axes (query API, semantic layer, extraction depth, CI substrate) — see
docs/why-codemap.md§ Codemap vs alternatives. Do not clone peer design in prose; reach for the underlying spec (plan-pr-inspiration-discipline).
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.
- yesterday First seen · 122 lines · 65 tokens per session scan A f0bf1ff19764
docs-voice is a skill published in the GitHub repository stainless-code/codemap (8 stars, last pushed 6d ago), licensed MIT. It adds 65 tokens to every session and 1,726 once invoked, about $0.0003 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
roam
Codebase comprehension via roam-code CLI. Use when exploring codebases, planning modifications, debugging failures, assessing PR risk, or checking architecture health. Triggers on: understanding project structure, pre-change safety checks, finding symbols/files, blast radius analysis, affected tests, health scoring…
map-learn
Capture reusable lessons after a completed MAP workflow. Use when a MAP run has finished and you want rules written to .claude/rules/learned/ from a workflow summary or handoff. Do NOT use during active implementation.
source-reading
带 AI 精读大型开源仓库，产出每句话都能回溯到源码具体行的书稿、课程或技术文档。核心是零幻觉：每一处引用逐字节核实、行号实读、静默删行由脚本抓出。覆盖锁版本锚点、写逐章大纲、八段结构写章节、编成带封面封底的 HTML 书、机器校验、并行子 Agent 生产六件事。当用户要读懂一个陌生的大型仓库、精读某个开源项目源码、把源码整理成一本书、整理成课程或系列文章、做源码解读、写架构分析、或者要派多个 Agent 并行写技术内容时使用。触发词：精读源码、读源码、源码解读、源码分析、拆解这个项目、这个仓库怎么读、把源码写成课、把源码写成书、写源码精读、架构分析、code walkthrough、带我读代码。.
levelup-specify
Extract Context Directive Records (CDRs) from the current session after completing work. Identifies reusable patterns (rules, personas, examples, evals) and captures directive compliance cases for team-ai-directives.
lane-memory
SMA-style project fact corpus under .agents/memory/. Opt-in via adoc stages.memory.enabled. Use when user says память, lane-memory, corpus, CORE, почему бот забыл, or an agent needs durable non-code facts. Not PROGRESS/LESSONS dumps.
t-doc
Generate project tutorial documentation that reads like a human wrote it, not like AI output. Scans the codebase, extracts architecture/APIs/config/deployment details, and writes structured tutorial docs in the project's documentation target, preferring an existing docs-web site over docs/tutorials/. Use this skill…