learning-curator

A learning-capture agent that records a specific mistake or surprise as a Markdown candidate file. It leaves broader rules and guides for a separate synthesizer.

In plain words
What is it for?
Use it after a review finding, incident, unexpected result, or other mistake. It records what happened, what was expected, the source, and the lesson.
Why use it?
It preserves useful lessons without mixing unrelated incidents or turning one example into a general rule too early.

Agent

Install

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.

agentmods
npx agentmods add agents/tacoda/keystone/learning-curator
Clone the repo
git clone --depth 1 https://github.com/tacoda/keystone
Per session 23 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 244 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00023 $0.00244
Opus 5 $0.00012 $0.00122
Sonnet 5 $0.00005 $0.00049
Haiku 4.5 $0.00002 $0.00024

Measured 2d ago against content hash 053efa106b22, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

learning-curator 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.

Origin

Copies of this mod

1 near-identical copy found in the catalogue:

.charter/agents/learning-curator.md · 43 lines

What it actually says

Learning curator

You convert a surprise (something the agent or the charter got wrong) into a learning candidate file under .charter/learning/inbox/. You do NOT promote it to corpus or guide — that's the synthesizer's job.

Posture

  • One candidate per file. Don't merge unrelated lessons.
  • Cite the artifact that triggered the surprise (commit, review, incident link).
  • No premature generalization. State the concrete instance; let synthesis decide the rule shape later.

Output

A markdown file at .charter/learning/inbox/<slug>.md with:

---
captured: <iso8601>
trigger: <commit | review | incident | other>
source: <link or path>
status: candidate
---

Followed by: what happened, what was expected, what we learned.

Changes

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.

  1. 2d ago First seen · 43 lines · 23 tokens per session scan A 053efa106b22

Subscribe to this mod's changes

learning-curator is an agent published in the GitHub repository tacoda/keystone (44 stars, last pushed 2mo ago), licensed MIT. It adds 23 tokens to every session and 244 once invoked, about $0.0001 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 2026-08-30.