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/chianw/c31/c31-context-engineeringnpx skills add ChianW/C31 --skill c31-context-engineeringgit clone --depth 1 https://github.com/ChianW/C31Wrote 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/chianw/c31/c31-context-engineering)<a href="https://agentmods.dev/skills/chianw/c31/c31-context-engineering"><img src="https://agentmods.dev/badge/skills/chianw/c31/c31-context-engineering.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 | $0.00040 | $0.02109 |
| Opus 5 | $0.00020 | $0.01055 |
| Sonnet 5 | $0.00008 | $0.00422 |
| Haiku 4.5 | $0.00004 | $0.00211 |
Grade A, and why
C31-context-engineering 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 4d 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 — 233 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Multilingual Triggers
| Language | Trigger phrases |
|---|---|
| EN | context, 上下文, 信息组织, 变蠢, context engineering, 信息太多, 你好像忘了, 上下文不够了 |
| ZH | 上下文工程, 管理上下文 |
| JA | コンテキストエンジニアリング, コンテキスト管理 |
Output language: Respond automatically in the user's conversation language.
C31 Context Engineering
The core pain point of the C31 system: a 200K context window, but with poor information organization = effective capacity of only ~30K.
This is not a capacity problem — it is an information structure problem.
Core Principles
- Layered loading — Not all information is worth loading
- On-demand retrieval — GBrain searches only when needed, not on every call
- Active discarding — Expired, completed, or irrelevant information is cleaned up promptly
- Anti-rationalization — Do not retain information "just in case"
Information Layering Architecture (Four-Layer Model)
| Layer | Content | When to Load | Discard Condition |
|---|---|---|---|
| L1 Immediate | Current conversation, task instructions for this turn, active tool calls | Always loaded | Discard when conversation ends |
| L2 Working | Current project files (STATE.md, PLAN.md, CONTEXT.md), active todos | Load at session start | On project switch or completion |
| L3 Rules | AGENTS.md, SKILL.md, SOUL.md, USER.md | Load on-demand at session start | Replace when rules are updated |
| L4 Memory | Historical knowledge, solutions, past projects, long-term memory | Search only when needed | Never proactively loaded |
Hard rule: L4 is never automatically injected into context. It must be explicitly retrieved via GBrain search or memory_search.
Rules Files Management
On-Demand Loading Strategy
AGENTS.md uses modular splits (core.md, decision-boundary.md, communication.md, workflow.md).
Loading rules:
- Every session → load
core.md(always) - Involves decisions/permissions → load
decision-boundary.md - Involves message format/platform → load
communication.md - Involves scheduling/cron/heartbeat → load
workflow.md - Sub-files not relevant to the current task are not loaded
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.
- 4d ago First seen · 233 lines · 40 tokens per session scan A c2a5fb1dac43
C31-context-engineering is a skill published in the GitHub repository ChianW/C31 (1 stars, last pushed 9d ago), licensed MIT. It adds 40 tokens to every session and 2,109 once invoked, about $0.0002 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
code-review-context
Codex maintains a context (history of messages) that is sent to the model in inference requests.
hive.context-preservation
Proactively extract critical values from tool results into working notes before automatic context pruning destroys them.
hive.note-taking
Maintain a free-form scratchpad of decisions, extracted values, and open questions so context pruning doesn't lose anything you still need.
enhance-claude-memory
Use when improving CLAUDE.md or AGENTS.md project memory files.
learn
Research any topic online and create learning guides. Use when user asks to 'learn about', 'research topic', 'create learning guide', 'build knowledge base', or 'study subject'.
perf-profiler
Use when profiling CPU/memory hot paths, generating flame graphs, or capturing JFR/perf evidence.