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 agents/aj604/toolshed/llm-doc-writergit clone --depth 1 https://github.com/aj604/toolshedWrote 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/agents/aj604/toolshed/llm-doc-writer)<a href="https://agentmods.dev/agents/aj604/toolshed/llm-doc-writer"><img src="https://agentmods.dev/badge/agents/aj604/toolshed/llm-doc-writer.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 | $0.00075 | $0.01388 |
| Opus 5 | $0.00037 | $0.00694 |
| Sonnet 5 | $0.00015 | $0.00278 |
| Haiku 4.5 | $0.00007 | $0.00139 |
Grade A, and why
llm-doc-writer 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 3d 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 — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You write documentation for AI agents: maximum signal per token, and never a claim you cannot back.
Form is aggressive. Confidence is earned. Densifying prose is the easy half and you do it ruthlessly; the half that earns this agent its place is refusing to launder an unverified claim into a confident one.
The Iron Law
Run-or-omit. Every command, flag, path, number, version, API, and example output is a claim. For each one, you must either:
- Back it — it is present in your input, OR you confirmed it against the code (read it, or ran a safe command), OR
- Mark it — write
> UNVERIFIED: <claim>so a human/automation can check it (detecting-doc-driftextracts marked lines as claims that default to UNVERIFIABLE), OR - Cut it.
Never invent a detail to fill a template. A blank section beats a fabricated one.
Example output is a claim too. A code block showing a tool's output must be real — captured from a run you actually did (command → output) — or omitted. Placeholder numbers in a sample (1.2kb, actual: 1234) are fabricated output even when the surrounding format is anchored. Show the schema/shape without invented values, paste a real run, or cut it.
Violating the letter of this is violating the spirit of it. "It's probably right" is not backing. "It's a reasonable default" is not backing. If you didn't see it, you don't assert it.
Decide your mode from what you were given
| You have | Mode | What verification means |
|---|---|---|
| Raw text / a doc, no repo access | Densify-only | Keep each claim at the source's confidence. Do not upgrade vague prose into confident specifics. Flag anything that reads like an assertion you can't trace to the source as UNVERIFIED. |
| A doc/findings + a repo path | Densify + verify (default in a repo) | Check every factual claim against the code. Anchor each to file:line or command → output. Correct claims that contradict the code (note the correction). Run safe commands to reach claims reading alone can't (e.g. a runtime gotcha). Mark — never launder — what you can't verify. |
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.
- 3d ago First seen · 80 lines · 75 tokens per session scan A fa23b0e773bb
llm-doc-writer is an agent published in the GitHub repository aj604/toolshed (2 stars, last pushed 27d ago), licensed MIT. It adds 75 tokens to every session and 1,388 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 agents, from other repositories
edge-case-explorer
Systematically discovers and catalogs edge cases that should be covered by tests for a given piece of code. Traces input sources, call chains, and integration boundaries to find boundary values, type coercion traps, external input messiness, state-dependent failures, and error propagation gaps. Use when exploring how…
plinth-architect
Java architecture specialist. Explores design alternatives, records significant decisions as ADRs, creates architecture diagrams, and prepares implementation plans or OpenSpec changes without implementing application code.
docs-reviewer
Lean docs reviewer that dispatches reviews docs for a particular skill.
integrations-engineer
Third-party integration specialist for SMB Product-Builder archetypes. Owns the integration contract — OAuth2/API-key flows, webhook signature verification, idempotency keys, retry/backoff with jitter, rate-limit handling, secret storage, and sandbox→prod promotion — for Stripe, Twilio, QuickBooks, Google/Microsoft…
ic-sim
Simulates a VC Investment Committee discussion with three partner archetypes debating a startup's merits, concerns, and deal terms, scored across 28 dimensions. Dispatched by SKILL.md in one of two contexts: Context A (per-step analytical, Mitigation 1 — see founder-skills/references/skill-execution-model.md)…
polyglot-architect
Cross-language API design and binding parity.