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/datacore-one/datacore/roadmap-buildnpx skills add datacore-one/datacore --skill roadmap-buildgit clone --depth 1 https://github.com/datacore-one/datacoreWrote 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/datacore-one/datacore/roadmap-build)<a href="https://agentmods.dev/skills/datacore-one/datacore/roadmap-build"><img src="https://agentmods.dev/badge/skills/datacore-one/datacore/roadmap-build.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 | $0.00111 | $0.04334 |
| Opus 5 | $0.00056 | $0.02167 |
| Sonnet 5 | $0.00022 | $0.00867 |
| Haiku 4.5 | $0.00011 | $0.00433 |
Grade A, and why
roadmap-build 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 today.
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 — 373 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Roadmap: build one agents can execute
Derived from rebuilding the PLUR roadmap, 2026-09-01 to 2026-09-03: five drifted documents to one validated source of 93 epics, 750 tasks, and three checks that catch regressions nobody would notice by reading.
The one thing that matters
A roadmap is good when it is CHECKABLE, not when it is well written.
Every structural finding in that rebuild came from a script, never from reading the file: five areas built and invisible, eleven epics whose repo state contradicted them, fourteen agent-doable tasks parked behind a wrong flag. The prose was fine throughout. The prose is always fine.
So: build the checks early and let them find the work. Do not write the whole roadmap and validate at the end.
Phase 1 — Find what is already true
Before writing anything, read the repos. Roadmaps decay by omission far more than by error, and the omissions are invisible from inside the document.
gh issue list --repo <org>/<repo> --state open --limit 200
gh pr list --repo <org>/<repo> --state open
git -C <repo> log --oneline -40
ls <repo>/src/<subsystem>/ # whole subsystems with no roadmap presence
Expect to find work that shipped and work that is invisible. In the PLUR rebuild: a provenance subsystem shipped in a repo the roadmap represented with two gated lines; eleven modules and three shipping tools with zero items; a 49-route admin surface with no coverage at all.
Ask of every subsystem: does the roadmap know this exists? Where it does not, that is an epic, and its status is whatever the code says — not what the document assumed.
Phase 2 — One writable source
Pick ONE file. Everything else becomes a generated view or a marked pointer.
Why fragmentation happens, and it is not carelessness: every document is writable, so ideas land in whichever one is open, and none is wrong enough to fix. The fix is not discipline. It is removing the ability to write.
Mark every other roadmap-shaped file with a canonical pointer naming what an item added there would be invisible to. Then add a check that flags a new unmarked one, because a sixth document is how the first five happened.
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.
- today First seen · 373 lines · 111 tokens per session scan A 417ecacccbdb
roadmap-build is a skill published in the GitHub repository datacore-one/datacore (4 stars, last pushed today), licensed MIT. It adds 111 tokens to every session and 4,334 once invoked, about $0.0006 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 2026-09-05.
Other skills, from other repositories
learn-from-correction
Propose a principle edit to a skill or persona file based on a (agentoutput, humanoutput) correction pair. Outputs a proposal markdown for human review - never auto-edits the target file.
council
Multi-persona debate for founder decisions. 4 personas argue a topic across structured rounds.
add-editor
Wire Deus's memory + evolution layers into an external code editor (Zed and other ACP/MCP clients) so the editor's own agent gains Deus's vault recall and self-improving reflexion loop. Configuration only — no new code. Triggers on "add editor", "use Deus in my editor", "zed integration", "editor integration", "wire…
quiz-me
Use right after writing or editing code or files for the user — before ending your turn — to quiz them on what was just built and confirm they actually understand it. Also triggers on "quiz me", "test my understanding", "check what I learned", "comprehension check", or recovering the understanding that vaporizes when…
add-llama-cpp
Install and verify a local llama.cpp server for optional Deus local-generation experiments. Keeps Ollama as the required default for embeddings and judge work.
code-review
Multi-agent code review with learning feedback loop — reviews PRs using parallel specialized agents (style, logic, security) with confidence scoring and false-positive reduction.