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 cyberchitta/llm-context.py --skill lc-curate-contextgit clone --depth 1 https://github.com/cyberchitta/llm-context.pyWrote 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/cyberchitta/llm-context.py/lc-curate-context)<a href="https://agentmods.dev/skills/cyberchitta/llm-context.py/lc-curate-context"><img src="https://agentmods.dev/badge/skills/cyberchitta/llm-context.py/lc-curate-context/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/cyberchitta/llm-context.py/lc-curate-context"><img src="https://agentmods.dev/badge/skills/cyberchitta/llm-context.py/lc-curate-context.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00090 | $0.01378 |
| Opus 5 | $0.00045 | $0.00689 |
| Sonnet 5 | $0.00018 | $0.00276 |
| Haiku 4.5 | $0.00009 | $0.00138 |
Grade A, and why
lc-curate-context 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 9d 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 — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Context Descriptor Creation
Three jobs, in order: decide what a task needs, verify the decision against the codebase, pack it for whoever is reading.
The first is not "gather a lot of relevant code" — it is to produce the smallest full-file and excerpted-file set that is still enough to finish the task. The second is what lc-preview is for, and it is not optional: a rule that looks right and omits the module its edit targets call will fail quietly. The third depends on the consumer, and getting it wrong is also quiet — see "Packing for a Sub-Agent".
Workflow
- Understand the task in concrete file terms.
- Pick the narrowest sensible baseline filter (see below).
- Put files you expect to edit in
full-files. - Put callers, dependencies, and large reference files in
excerpted-files. - Run
lc-preview -r <rule>. Read the exactFull files/Excerpted fileslists, thenReferenced but not selected. - Tighten until the selection is minimal and sufficient.
- Generate the real output once before handing the rule off —
lc-context -r <rule>— and confirm it contains file bodies, not just a tree.
Pick the Baseline First
Start narrow unless you have a reason not to.
| Baseline | Use when |
|---|---|
lc/flt-no-files |
Surgical task; you know the likely files; you want exact control over membership. Default for task rules. |
flt-repo-base or lc/flt-base |
Broad project slice; the task spans a subsystem; the repo already has a disciplined baseline filter. |
If lc-preview shows unexpected full files, the baseline is too broad.
Full vs Excerpted
Full — files you expect to edit; small configs or templates that control the behaviour; compact integration points where exact code matters.
Excerpted — callers and dependencies; large modules where structure is enough; reference implementations and surrounding architecture.
Move a file from excerpted to full only when the exact body matters.
Verify with lc-preview
What ships with it
5 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.
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.
- 9d ago First seen · 108 lines · 90 tokens per session scan A 9c5a564c483b
lc-curate-context is a skill published in the GitHub repository cyberchitta/llm-context.py (306 stars, last pushed 13d ago), licensed Apache-2.0. It adds 90 tokens to every session and 1,378 once invoked, about $0.0005 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.
Other skills, from other repositories
MCP Integration Assistant
Helps design and implement Model Context Protocol (MCP) server integrations for AI agents.
MCP Server Builder
Step-by-step guidance for creating a new MCP server with FastMCP or the TypeScript SDK — tool definitions, resource handlers, error responses, and usage examples.
crystalline-capture
Use when durable knowledge is learned while working, when the user asks to update a Crystalline domain, or before writing or editing an engram so it is deduplicated, well-formed and linked correctly.
crystalline-intelligence
Use when the crystalline MCP tools are available and the task involves recalling, storing or sharing knowledge - searching domains, capturing engrams or sharing with a team.
wellread
Collective research memory for AI coding agents — search prior research before hitting the web, save findings back to the shared cache, view karma stats. Use when the user asks a library/API/framework question, before running WebSearch/WebFetch, or after deep web research so the next agent doesn't repay the cost.…
structured-memory
Use whenever persistent agent memory is read, written, or has grown past its budget - the first time memory is consulted in a session, whenever something worth remembering appears, and whenever MEMORY.md or a topic file exceeds its size budget. Implements a two-tier structure (slim MEMORY.md index linking memory-.md…