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 skills add yasik/kmd --skill kmd-ingestgit clone --depth 1 https://github.com/yasik/kmdWrote 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/yasik/kmd/kmd-ingest)<a href="https://agentmods.dev/skills/yasik/kmd/kmd-ingest"><img src="https://agentmods.dev/badge/skills/yasik/kmd/kmd-ingest/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/yasik/kmd/kmd-ingest"><img src="https://agentmods.dev/badge/skills/yasik/kmd/kmd-ingest.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 | $0.00135 | $0.02377 |
| Opus 5 | $0.00068 | $0.01189 |
| Sonnet 5 | $0.00027 | $0.00475 |
| Haiku 4.5 | $0.00014 | $0.00238 |
Grade A, and why
kmd-ingest 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 8d 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 — 200 lines — stays where its author put it; the contents beside it link to each section on GitHub.
kmd-ingest — the KB write protocol
The knowledge base compounds in value only if every write follows the same discipline: condensed pages, declared provenance, cross-references swept, one log entry per operation. A single undisciplined write is cheap; a thousand of them make the KB untrustworthy at query time. This skill is that discipline — the same whether the KB belongs to one person or an organization of agents. Permissions are open — any author may create or edit any page. The manner is not: every write, down to a one-line correction, is an ingest.
The two layers (get this right and the rest follows)
sources/— append-only evidence. Raw external material: scraped pages, whitepapers, transcripts. You may add files here (when new raw material is involved in your work); you never edit or delete existing ones.- Everything else in the KB — living synthesis. Pages you write and
maintain (
entities/,concepts/,projects/, …) plus derived artifacts inassets/(charts, decks — embedded from pages, never pages themselves).
Your own outputs — reports, analyses, lessons — are never filed into
sources/, no matter how polished or reference-like they feel. Synthesis
recycled as evidence is how hallucinations launder themselves into ground
truth. A finished report becomes a page citing the raw material it was
built from.
Before you write
- Locate the KB root — in order: a path you were given;
$KMD_ROOT; a.kmd.jsonat the workspace root ({"root": "<dir>"}); a directory containingschema.md/log.md; the defaultkb/under the workspace. The bundled scripts implement exactly this resolution (--kbflag). - Pick the transport — read
.kmd.json'stransportkey; if missing, runpython3 scripts/detect_transport.py --kb <kb-root>once. When it saysobsidian-cli, read and write pages through the Obsidian CLI per references/obsidian-cli-transport.md; otherwise (or whenever a CLI call fails) use plain filesystem tools. - Read
schema.mdat the KB root — the single source of truth for taxonomy, writing style, and conventions; its Extensions section may add page types and hierarchies beyond the defaults, and it overrides anything here that conflicts. If the KB has no schema.md yet, bootstrap it from references/schema-template.md — stripping the<!-- kmd:template-note -->block, and the org-extension block unless this is an org installation. - Check for the org extension — if
.kmd.jsondeclares an"org"key or the workspace has a charter with anORG.md, read references/org-extension.md first (it adds ownership routing and org-specific provenance origins). Otherwise ignore.
What ships with it
8 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.
- references/obsidian-cli-transport.md 3.6 KB
- references/org-extension.md 3.0 KB
- references/schema-template.md 7.7 KB
- scripts/detect_transport.py 4.7 KB runs code
- scripts/kb_common.py 22 KB runs code
- scripts/kb_log.py 4.7 KB runs code
- scripts/recompile_index.py 4.4 KB runs code
- scripts/validate_page.py 2.4 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.
- 8d ago First seen · 200 lines · 135 tokens per session scan A bceb11f4a420
kmd-ingest is a skill published in the GitHub repository yasik/kmd (2 stars, last pushed 1mo ago), licensed MIT. It adds 135 tokens to every session and 2,377 once invoked, about $0.0007 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
karpathy-llm-wiki
Use when building or maintaining a personal LLM-powered knowledge base. Triggers: ingesting sources into a wiki, querying wiki knowledge, linting wiki quality, 'add to wiki', 'what do I know about', or any mention of 'LLM wiki' or 'Karpathy wiki'.
wiki
Operate on the user's personal Obsidian LLM-Wiki vault (repository root). Use when the user invokes /wiki, asks a question that should be answered using their knowledge base, asks to ingest a new source into the wiki, asks to lint or health-check the wiki, or asks to generate a document (analysis, comparison, summary…
link-memory
Use after important user-approved decisions, when durable context should be proposed or reviewed, and for explicit Link memory lifecycle work: remember, recall, review, update, archive, restore, forget, or explain local memories through the CLI without requiring MCP.
link-retrieve
Use before answering work that may depend on user memory, project history, source-backed notes, or prior decisions; retrieve compact Link context through the CLI without loading the whole wiki or requiring MCP.
link-ingest
Use when raw files are present, source pages look stale, or a user asks to ingest notes into Link; refresh source-backed wiki pages, propose memories, and validate updates through the CLI without MCP.
lean-ctx
Local context tooling for AI agents. Use it to select, shape, reuse, recover, and inspect context before inference when reading files, running shell commands, searching code, or exploring directories.