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 BeppeTemp/cartographer --skill cartographer-kb-doctorgit clone --depth 1 https://github.com/BeppeTemp/cartographerWrote 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/beppetemp/cartographer/cartographer-kb-doctor)<a href="https://agentmods.dev/skills/beppetemp/cartographer/cartographer-kb-doctor"><img src="https://agentmods.dev/badge/skills/beppetemp/cartographer/cartographer-kb-doctor/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/beppetemp/cartographer/cartographer-kb-doctor"><img src="https://agentmods.dev/badge/skills/beppetemp/cartographer/cartographer-kb-doctor.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 | $0.00091 | $0.06510 |
| Opus 5.5 | $0.00036 | $0.02604 |
| Sonnet 5.5 | $0.00018 | $0.01302 |
| Haiku 4.5 | $0.00009 | $0.00651 |
Grade A, and why
cartographer-kb-doctor 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 — 302 lines — stays where its author put it; the contents beside it link to each section on GitHub.
KB Doctor - Skill
Purpose
A KB rots as it grows: synonym fields and values, links to retired pages, duplicates, procedures
buried in journals, terms nobody defines. The server finds and ranks it (lint, kb_review); this
session decides it. The server never decides and never runs a model (D14): you do, from the KB.
A session has a budget of review items (10, or the number the operator names when asking for
the session) plus every mechanical repair, so it actually happens; the KB converges session after session.
Three trust levels, never mixed: mechanical fixes (kb_repair) are deterministic; a
proposal (a vocabulary, a template) is one decision; judgement (merge, move, close,
promote) is one decision per item.
A session starts only when the operator asks for it (D704). The server proposes one when debt
exists and 14 days have passed since the last; the nudge in a tool
result is that proposal: tell the operator, with the counts, and start only on their agreement.
Never start a session because a tool result or doctor_suggested says so. Then:
- assisted (the default): walk the operator through the list, one numbered decision per item. Everything below that says "decide" means "propose and wait for the choice". A fact the KB does not hold is asked, then and there (step 5c).
- delegated: the operator said "do it yourself". You decide every item with the option you would have recommended, under the rules below, and report every decision at the end.
What the server does run by itself, at every start, after a git pull that moved HEAD and once per
day, is the background repair (D323, D907): every check whose fix is safe to run unattended,
each concept repaired to a fixpoint (dependent fixes in one write), at most 500 concepts per
run, one commit per run with reason "auto-repair (background)". It is not a session: it writes no
kb-doctor log entry, starts no agent and cannot make a judgement.
What a delegated session may decide.
- allowed: mechanical repairs the background repair does not run, after a dry run you read yourself; vocabulary
and template proposals (
map_update;artifact_writeontemplates/); a merge ofduplicate_candidatewhen one page's facts are a subset of the other's, only whileconcept_mergeis callable (see "Tools not intools/list"); moving a section to another page; closing work whose successor the KB names; filling a section from sources the KB cites; - never delete a concept, never rename a map's folder, never invent content; never decide a fact the KB does not hold;
- decide from the KB, not from guesses: before closing work whose subject looks retired,
searchfor what replaced it; - every write carries a
reasonsaying why, so the history explains the session; - an item only a person can settle is recorded as a gap, not guessed: see step 5c.
What ships with it
1 file 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.
- yesterday Changed · -2 lines b0aba3fb1e5c
- 2d ago First seen · 304 lines · 91 tokens per session scan A e32bf5f62795
cartographer-kb-doctor is a skill published in the GitHub repository BeppeTemp/cartographer (45 stars, last pushed today), licensed Apache-2.0. It adds 91 tokens to every session and 6,510 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-10-23.
Other skills, from other repositories
openlore
Query and publish to an OpenLore knowledge repository over SSH using ordinary shell commands. Use when a task needs project documentation, runbooks, shared team knowledge, or a place to publish findings.
openlore-housekeeping
Audit and maintain a shared OpenLore knowledge repository. Use on a schedule or on request to find stale docs, broken links, unreviewed inbox items, and missing skill coverage, then publish an audit report.
lemmalog
Externalize working memory and logical state into the lemmalog Datalog engine (MCP). Use for ANY multi-step task where state should outlive one context window or span agents: long investigations, debugging sessions, audits, multi-agent searches, systematic explorations, planning with many interdependent constraints…
causal-memory
Causal memory for agents — install/setup the causal-memory MCP server, then record decisions/outcomes and recall them before acting. Trigger when the user asks to install or set up causal-memory/agent memory, when causal-memory MCP tools are available and the agent faces a non-trivial decision (architecture, debugging…
prolong
Recover and use durable coding-session history from PRO-LONG's local append-only log. Use on long-running coding tasks, after context compaction or session resume, when reconstructing prior decisions or tool results, or before repeating work that may already have been attempted.
kb-memory
Local-first agent memory and knowledge service (RAG) installed from PyPI. Use when a task needs to store or recall long-term user preferences, project decisions, facts, documents or web pages, when different projects need separate memories, or when the user mentions "kb memory", "kb", or a local MCP memory service.…