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/n-n-code/n-n-code-skills/documenternpx skills add n-n-code/n-n-code-skills --skill documentergit clone --depth 1 https://github.com/n-n-code/n-n-code-skillsWhat 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 | $0.00104 | $0.01600 |
| Opus 5 | $0.00052 | $0.00800 |
| Sonnet 5 | $0.00021 | $0.00320 |
| Haiku 4.5 | $0.00010 | $0.00160 |
Grade A, and why
documenter 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 2d 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 — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Documenter
Boundaries And Composition
- Treat review and chat-only draft requests as read-only. Edit repository files only when the user asks to create or update them. If the user later switches to apply mode, refresh the target, evidence, applicable guidance, and initial diff before editing.
- Before editing, inspect the target and relevant initial diff when available. Preserve and attribute existing changes, patch around clear overlaps, and stop rather than overwrite when ownership or intended content is ambiguous.
- Use
agents-md-generatoras primary for root or nested repositoryAGENTS.mdfiles. - Add
documenter-coauthoringfor user-directed context gathering, outline agreement, or staged drafting. Document length neither selects nor excludes it. - When implementation, release machinery, or doc-site infrastructure also changes, use the relevant implementation or project skill for that work and apply this skill to the documentation content.
- For an authorized change involving generated output, find its authoritative source, determine whether source and output must remain synchronized, and inspect the generator's side effects before editing. Regenerate only when safe and authorized; never hand-edit the output. If required synchronization cannot be completed, do not create a partial inconsistent state. When local policy permits temporary divergence, leave it only with explicit user acceptance and never report that state as complete. If a failed regeneration already created a partial state, undo only this task's safely attributable delta when possible; otherwise stop. When synchronization remains blocked, report exact source and output states, the blocking side effect or dependency, and the smallest authority or safe tooling needed to continue.
- Use a specialized workflow for regulated document control.
Core Workflow
- Classify the work before writing:
- intent: review, chat-only draft, or apply
- audience and reader task
- document type and output target
- claim status: current, proposed, normative, or historical
- Read the target, applicable local guidance, existing changes, and evidence
relevant to affected claims:
- current behavior: code, config, schemas, tests, safe command output, shipped artifacts, and applicable versioned primary sources for external contracts
- proposed behavior: supplied requirements, draft decisions, explicit user direction, and visible open questions
- normative behavior: accepted decisions, approved specifications, policies, and standards
- historical claims: versioned records, changelogs, release artifacts, ADRs, and other dated evidence Local guidance controls placement and conventions; claim-specific evidence controls factual content. Reconcile conflicts by scope, version, and authority. Expose unresolved material conflicts and pause when the requested edit requires choosing a source; never choose silently or change implementation to make prose true. A proposal may preserve an unresolved conflict, but factual current-state documentation must label a disputed material claim as unknown or stop when settling it is required. For review-only work, validate as applicable and return prioritized, evidenced findings with limitations; skip drafting and edit-only completion. For a chat-only draft, continue without repository writes and label its validation status.
- Choose the smallest coherent change:
- update only affected sections for a narrow edit
- outline a new or large document before drafting it autonomously
- before restructuring, map meaningful content as retained, moved, or intentionally removed and search for inbound links or anchors that may break; when applying the change, update in-scope callers, preserve a compatible anchor, or defer the breaking removal and report it
- Draft in the repository's native markup and language conventions:
- organize standalone pages around one dominant reader task; gateway documents such as READMEs may deliberately orient and route several tasks
- mark planned behavior and illustrative examples clearly
- keep examples free of credentials, private data, and environment-specific secrets
- for API comments, document contracts and non-obvious behavior rather than narrating syntax
- Validate proportionally:
- inspect validation commands for file writes, generated output, dependency installation, and external effects before running them; in review or chat-only draft mode, use read-only or check-only variants
- run applicable existing documentation builds, formatters, linters, or link checks
- check local paths, anchors, identifiers, filenames, and cross-references
- run examples only when they are safe and local
- do not publish, deploy, release, install dependencies, migrate data, or cause other consequential mutations without the required authority; authoring or validation intent does not supply that authority
- report important examples and commands as observed, statically inspected, illustrative, or unverified when the distinction matters
- Tighten and hand off: remove duplication and accidental placeholders, link new durable docs from the nearest existing index when appropriate, review the resulting diff when files changed, and report checks plus material uncertainty.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 2d ago First seen · 138 lines · 104 tokens per session scan A 720a8f760ae6
documenter is a skill published in the GitHub repository n-n-code/n-n-code-skills (4 stars, last pushed 4d ago), licensed MIT. It adds 104 tokens to every session and 1,600 once invoked, about $0.0005 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
link-ticket-to-session
Link the current Claude Code session to a ticket (Linear, Jira, GitHub Issues, or GitHub Pull Requests) and cache its title/status in karma. Use when the user explicitly asks to link, attach, associate, or connect this session to a ticket, issue, or PR — e.g. "/link-ticket-to-session ABC-123", "link this session to…
laundering
Use when all territories report completion after the hit — handles integration, verification, consigliere final review, cleanup, and ledger updates to produce clean production-ready code.
reconnaissance
Use when beginning a new heist — deploys associates to survey the target codebase, existing tests, dependencies, documentation, and the ledger to produce a reconnaissance dossier for the don's review.
the-grilling
Use when adversarial brainstorming is needed after reconnaissance — the Proposer, Devils-Advocate, and Synthesizer each run ONE pass (no rounds); the Don is grilled on the IDEA first, one question at a time, then the proposal is attacked.
using-gangsta
Use when starting any conversation — establishes how to find and use Gangsta skills, enforces the 1% invocation rule, maps platform tools, and routes the Don's intent through the Gangsta Agents Family hierarchy.
drill-tdd
Use when implementing any feature or bugfix — enforces the Red-Green-Refactor drill with no production code allowed without a failing test first.