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/backchainai/backchain-plugins/docsnpx skills add backchainai/backchain-plugins --skill docsgit clone --depth 1 https://github.com/backchainai/backchain-pluginsWhat 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 | $0.00173 | $0.03914 |
| Opus 5 | $0.00086 | $0.01957 |
| Sonnet 5 | $0.00035 | $0.00783 |
| Haiku 4.5 | $0.00017 | $0.00391 |
Grade A, and why
docs 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 — 253 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Docs
Route one document to the mode it belongs in, write it under that mode's constraints, then check its markdown form.
The unit of work is one document and one next action. This skill does not plan a documentation set top-down, does not audit a whole docs/ tree in one pass, and does not produce an outline for a future set of pages.
Two paths share the same routing step: writing a new document (the write path) and assessing one that already exists (the place path). Both start at the compass below, and both end at the same markdown-form check.
Where this does not apply
Diataxis governs product documentation written for a user. It does not govern everything that happens to be a markdown file in a repository.
Check the request against this table before routing anything:
| Out of scope | Why |
|---|---|
| ADRs and decision records | A dated historical artifact, not one of the four modes |
| Commit messages, PR descriptions | Not user documentation |
| Changelogs, release notes | Chronological by nature; no mode applies |
| Code comments and docstrings | Belong to the code, not the documentation set |
CLAUDE.md, agent and tool config |
Instructions to a machine, not a user |
When the request names one of these, say which row applies and stop before opening the compass.
None of the four modes fits an artifact that is fundamentally dated, historical, or machine-facing.
READMEs are in scope: advise on README structure when asked (trim it, reorder it, link out more aggressively), but never propose extracting its parts into separate mode documents the way this skill would for a contaminated tutorial or how-to guide. A README is a landing page and legitimately hybrid.
When scope is ambiguous
- A design doc that reads like project history belongs with ADRs, even if nobody labeled it one.
- A "docs" folder that turns out to hold internal runbooks for the team, not instructions for a product user, is out of scope until the request is reframed around an external reader.
- A file named
NOTES.mdsitting next to code is closer to a code comment than to any of the four modes; ask what reader it serves before routing it.
What ships with it
12 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.
- evals/evals.json 16 KB
- evals/trigger-evals.json 4.9 KB
- references/commonmark.md 8.4 KB
- references/explanation.md 5.6 KB
- references/frontmatter.md 5.0 KB
- references/how-to.md 5.5 KB
- references/reference.md 5.6 KB
- references/tutorial.md 6.0 KB
- scripts/check_markdown.py 29 KB runs code
- scripts/fixtures/clean-fenced.md 1.0 KB
- scripts/fixtures/malformed.md 684 B
- scripts/test_check_markdown.py 35 KB runs code
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 · 253 lines · 173 tokens per session scan A ec86b859aa22
docs is a skill published in the GitHub repository backchainai/backchain-plugins (4 stars, last pushed 26d ago), licensed Apache-2.0. It adds 173 tokens to every session and 3,914 once invoked, about $0.0009 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
diagnose-with-memory
과거 실패 기억을 회상해 버그를 진단하고 근본원인을 다시 기억에 남길 때 사용한다.
tdd-with-memory
과거 결정·컨벤션을 회상해 테스트 우선으로 구현하고 새 컨벤션을 기억에 남길 때 사용한다.
a-unslop-code
Finds what makes source code read as AI-written and points you at the parts that actually ship bugs. Sorts every tell into three buckets and fixes them in that order: bugs (swallowed errors, a made-up API, a left-in "rest of the code" stub), substance (tutorial-shaped boilerplate, over-engineering, code that ignores…
close-session
End-of-session ritual — audit today's patterns against accumulated memory, propose promotions, refresh MEMORY.md, and write the session handoff. Use when the user says "/memory-kit:close-session", "закрой сессию", "закрываем", "we're done for today", "wrap up".
innate-memory
Innate procedural knowledge layer for coding, debugging, and analysis sessions. ACTIVATE when: (1) user says "remember", "save", "log", "follow this rule", "don't forget", "recall", or "what did we learn"; (2) a non-obvious solution, workaround, constraint, or invariant was discovered during this session; (3) starting…
kayba-stage-5-action-plan
Triage each insight into discard/code-fix/prompt-fix and produce a prioritized action plan with specific recommendations. Trigger when the user says "run stage 5", "make action plan", "triage skills", or when invoked by the kayba-pipeline orchestrator. Requires eval outputs from stages 1-4.